Comps outlines

Ethical leadership

(Brown) What is it? -Honest, caring, principled leaders who make fair decisions -vs TFL: More transactional. You emphasize morals, not vision. Cognitive? -Social learning/modeling (Bobodoll-style) Individual diffs? -Probably conscientiousness How does it work? - Multi-level cascade of ethical climate across all levels of the organization - Creates team ethical culture How do you get more of it? -Face intense moral issues. Cites * Schaubroeck et al.

Charismatic leadership

(Hater & Bass) - Just charisma component of TFL. -(TFL brings together multiple styles.) -Arises from network centrality (doesn't cause centrality) --Sacrifice, vision

New school leadership styles

(Lord et al. 100yrs) 1. TFL 2. Transactional (Burns & Bass) - Contingent rewards 3. Charismatic 4. Ethical VERDICT: Different styles to seem to predict differentially, though TFL and Trans both good for performance. TFL augments other styles.

Missing data

(Probably won't have this question) Listwise: complete.cases() Pairwise: You only delete the case when it's needed for analysis. Like if you're looking at values and you're missing value data. But you still include that guy when looking at his personality relationship, which he has. Multiple imputation: -You regress Mike's GPA on time and you inject error (makes it stochastic),which gets you a diff estimate every time you simulate it, then you run that a bunch, and average the value you get. -Use stochastic regression to predict what I think the GPA is going to be for the semester you didn't report it. -Stochastic regression injects some randomness to simulate error for realism. Maximum liklihood: Picture: Simulated mean.........GRE score IT'S ABOUT ESTIMATING MISSING A CENTERAL PARAMETER LIKE A POPULATION MEAN, NOT ABOUT FILLLING IN LIKE ONE MISSING FIELD VALUE. You're trying to get the average GRE score, even though I I don't have the mean I just have sample values. You plugin in the GRE scores you have. It says "If there's mean of 1200, there's an X probabily of producing these scores, etc."

Job performance

(Theory from Campbell & Weirnick; outcomes from either Hunter & Schmidt or Delal or MB) What is it? Actions that contribute to org's goals (Campbell) *Models of perf?* 1. Task vs. Contextual (converting vs. replinishing) 2. OCB //is// a model of performance. 3. CWBs 4. Griffen: 3x3 Level (individ/group/org), adaptability, practivity 5. Competency models How to improve it? 1. Selection seems to impact it more than intervenable things. 2. Interventions still can have a significant impact according to a utility analysis by (Hunter & Shmidt). Especially *performance feedback* (with avg correlation of about *.4*, which is higher than most selection and attitudinal predictors) 3. Job attitudes can also impact it (Delal et al meta). Top predictors are... -OCBs & CWBs -Negative affect -Job satisfaction -Commitment, engagement, positive affect (in that order) *Nuance?* -Important to distinguish between performance itself and antecedents/outcomes of it (e.g., perf is not productivity and it's not org stock price).

Organizational justice

*(Colquitt meta for most of this)* What is it? -The socially constructed view of what is fair. -Distributive -Procedural -Interpersonal -Informational: Adequacy and truthfulness of explanations given. Underlies: -LMX -CWBs *Sell it?* -Interpersonal and informational justice has largest effect on evaluation of authority. Get more of it? -Leventhal criteria for PJ: --Consistence --Bias suppression --Accuracy --Correctability --Representativeness --Ethicality Nuance? -Justice dimensions highly related, but do offer incremental validity (Colquitt meta) -Process control well replicated factor in legal fairness (I can pick my own arguments and have time to make my case) *Future directions?* Colquitt recommendeds... -Justice interaction effects -More focus on interpersonal/informational justice Cites? -Adam's equity theory (inputs v. outputs) -Procedural justice: Leventhal -Justice climate (Colquitt) creates commitment -Shneider: 100yrs climate -Colquitt: Review of justice

range restriction

*How does range restriction affect your correlations?* It deflates them. *How do you correct for it?* 1. If you have the population standard deviation, you use Pearson's (1903) formula to weight the observed correlation by the population standard deviation. 2. If you don't know the population standard deviation, then you can use Cohen's 1959 formula to basically compare your data to a normal distribution and detect the truncation point, then generate a new mean and SD based on what your data would look like if it was normal. *How do you get the population variable?* You have to get access to some normative data. Like a dataset for DEVELOPING the IQ test that you directly range restricted by testing only people who passed the cutoff.

Frameworks for understanding org context?

*Setting* CAPTION (Tay et al) Complexity Adversity Positive Valence Typicality Importance Humor Negative Valence *Values* Competing Values Framework (Cameron & Quinn) Axes: Internal/integration focus VS external differentiation Flexibility VS. Stability MACH - Market, Adhocracy, Clan, Hierarchy

Survey design

*What is DIF (differential item functioning), why is it bad, and how can you prevent it?* -Can lead to incorrect conclusions about relationships.

Organizational politics

*What is it? (Chang et al. meta)* -Positive and negative advancement of self or ingroup-interests above others. -Positives: People with political skill have bigger networks and get get more resources for underlings. -Negative: Favoritism-based pay, coalition building). *How does it work?* (Chang et al meta) -Affect of politics on outcomes mediated through job attitudes. -So if work is hella meaningful that might buffer negative impact of org politics. -Tanks morale. Antecedents? -Minority status (Parker) Outcomes? -Lower performance and strain (Chang et al meta) -Can lead to positive outcomes when viewed as opportunity stress (Schuler) Best practices? Future directions? -Physically assess strain

Cross-cultural psychology

How does it work cognitively? * Enculturation: Learning a new culture * Acculturation: Deciding whether to change to a culture --ASMA *How do you sell caring about it to executives?* 1. Predicts emotions and attitudes better than IQ (Teras) *How do you measure it?* 1. Values: Schwartz's Universal values 2. Values: Hoffstede's cultural values 2a. Most predictive: Collectivism vs. Indivdualism 2b. Most predictive: Figure-ground 3. Eco/Political: Berry's ecological framework (environ vs. socio-political) What cites do you need to know? * Schwartz * Hoffstede * Berry

Organizational development

*What is it?* -A method of facilitating change in people, technology, and within organizational systems and processes with the goal of optimizing human fulfillment and task accomplishment. (Friedlander & Brown) -Research seeks to understand how change works and how to do it better (Alderfer) *Key concepts:* 1. *Orgs are systems.* You can't change one part without changing all parts (Katz & Khan). 2. *Freeze-move-refreeze* (Coch) 3. *Single vs. double-loop learning.* Single: Hit this one goal. Double: Hit this goal, or a different goal via a different method if you think that's more wise. 4. *1st/2nd/3rd-order change* (Bartunek & Moch 87) 5. *Punctuated equilibrium model* (Gersik). System will maintain equilibrium until the underlying structure stops supporting balance, then it will change far enough to reach a new equilibrium. 6. *Readiness for change* (Prochaska): An individual-level phenomenon. The extent to which change is possible and the target of change is desirable. 7. Absorptive capacity. *How do you get more readiness to change?* (Ford & Foster-Fishman) -Resources create Readiness: Orgs who have greater capacity for change seam to have more readiness. -People who've been working a while at an active job in a flexible, problem-solving work environment have max readiness for change. -Networks absorb Knowledge *Outcomes* -Readiness does relate to success of change effort (Jones) *How do you measure it?* Holt's readiness to change measure (change is needed, feasible, would benefit me, would benefit org, and management supports it). *Future directions* (Ford & Foster-Fishman) -Multi-level -Creating change readiness.

Team Processes

*What is it?* -Cues, meaning, planning, monitoring, communication, backup beh, leadership *How measured?* Most measured as emergent states: Tell me via self-report after what kind of happened before. Cites? -Burke *Future?* -Koz: More interventions on team process.

Training

*What is it?* -It's change: in behavior, knowledge, kill, and/or attitude How to measure it? - Intent to transfer - Multiple measures of on-the-job transfer bound to discrete, observable variables that are context-bound. *Sell it?* * Training IS change. Change is training. * Positive and significant with performance, profitability, and job satisfaction * Note: Perceived effectiveness higher than actual effectiveness *Best practices* 0. Needs analysis X. Secure support from management for training/transfer resources. X. Teach teachers transfer X. Provide opportunities for pre-training framing. --Tailor to existing schemas X. Expect transfer: Tell people you expect on-the-job transfer application X. Motivate to apply X. Build target behaviors into training program 4. Provide instruction aids/tools that convert well to job aids/tools. X. Demonstrate application X. Require a project X. Create feedback system for transfer X. Use multiple measures of transfer -- Observable, discrete change in behavior bound to a context --. Target changes that have relevant organizational impact 4. Booster sessions *Barriers* -Has to fit with The Dell Way ---Overcome 'training has to fit in with the Dell way' by assessing attitude towards the field. -Has to fit with prior beliefs about profession (doctors and meditation). --Fix: Prime a healer identity first, show evidence *Future directions?* 1. Closed skills -> Open skills 2. Factors outside training itself 3. Enhancing transfer 4. Time/change as dynamic/non-linear Cites: -Howto: Yelon & Ford -Baldwin, Blume, Ford -Robinson & Robinson (setting criteria) -Barriers: Roulston & Choi -Outcomes: Tharenou meta

Cognitive ability

*What is it?* -The acquired repertoire of all skills and knowledge available at a point in time (Humphreys) *Cognitive?* -Thickness of myelin sheath -Configural structure of the brain -Working memory *Individual differences?* -Evidence for both a general factor and specific cognitive abilities that are still predictive Sell it? -The single most predictive individual diff variable of job performance -Also valuable for training performance *How to get more of it?* -Education -Maybe Jeggi's working memory training -Maybe Lion's mane mushrooms Future? -Tailoring the narrow cog ability to the job Cites -Spearman: Created G and "some misc other uncorrelated specific factors" -Gilford: 150 abilities -Carroll: Stratum model, including gF/gC -Hunter & Shmidt: Outcomes of selection -Nye et al: Predictive validity of narrow abilities matched to job.

SJTs

*What is it?* -The lowest-fidelity simulation *Sell it?* -.18 with job performance (.35 for encumbants) *How do you do it?* -Job analysis with SMEs -Combine with theoretical models *Decision points* -Should do vs. would do framing? -Multimedia or text based? *Gremlins?* -Adverse impact -Ethical issues of SJTs in selection -Validity generalization -Ongoing validation *Best practices?* -Frame as should do -Multimedia is better than text, and reduces adverse impact. *Future directions?* -VR *Cites?* -Pollard: best practices -McDaniel & Campion: Meta with performance

Organizational culture

*What is it?* -The shared values and assumptoins that explain why organizations do what they do and focus on what they focus on. -Pr conscious 3 levels: 1. Artifacts 2. Espoused values 3. Assumptions and beliefs Cognitive roots? -Homophilly (ASA) -Fit -Values -Norms -Conformity Individual differences? Sell it? How do you manipulate it? -ASA -Stories & legends -Policy -Onboarding/mentoring How is it measured? -Competing Values Framework -Look for obvious artifacts and anomilies, listen for traumas -Survey to get values, norms with predefined dimensions -Ethnographic observations -Longitudinal change analysis -Ask about traumas; they prevent change. Cite? -Schein 1990 -Schneider et al. 100yrs of JAP Future directions? -Subcultures within orgs -Integration of climate and culture

Leadership (major theories)

*What is it?* -Winston: Influence others towards goal -Right person, right time. *Major theories (General cite: Lord et al. 100yrs)* 1. Wave 1: Great man (Bernard) - Every person has leader and follower traits. Leader sets collective goal and gets followers to commit to it. VERDICT: Some traits are common. 2. Wave 2: Behaviors - Initiating Structure & Consideration (Ohio state). VERDICT: Strong effects for both on org outcomes. 3. Wave 3: Contingencies/Situation (Fielder). LPC model: Relationship-oriented leaders describe bad workers more kindly than task-oriented leaders. 4. Wave 4: Confirming stuff with meta. 5. TFL, Gender, team. *Sell it?* (Hogan et al)* Consistent evidence for relationship between who's in charge and how well org performs. * From studies of CHANGES in leader (succession) *How do you measure it?* - Better: By the performance of the group they lead (Hogan) - Common: Survey people around the leader What cites do you need to know? * LMX (Graen) * TFL (Bass) * Team leadership (Marks, Zaccaro et al) * Charismatic leadership (House)

Job Satisfaction

*What is it?* A pleasurable or positive emotional state resulting from the appraisal of one's job or job experiences. (Locke) Value-precept theory (Locke): Job Satisfaction = (Want - Have) X Importance *How does it work cognitively?* * Evaluative process * Socially constructed *What individual differences effect it?* Straw and Ross found a ~.2 correlation between within-person job satisfaction ratings over a 5 year period, despite changing jobs and occupations. Thoresen & Judge - Meta found medium-to-large correlations between positive affect and high job satisfaction, and negative affect and low job satisfaction. Judge et al. - Core self-evaluations have medium-to-large correlations with job satisfaction, in theory because they co-opt the same evaluative process used in job satisfaction. Thibaut & Kellly - CL alt model formalized what Locke was saying in the first place. It might also be possible that positive, stable people with good self esteem/efficacy attain better jobs. *How do you sell caring about it to executives?* - Number one predictor of turnover (.70) - It is associated with higher performance in small amounts (Judge) *How do you manipulate it?* * Autonomy * Control over environment * Plenty of job resources * Select people with trait high job satisfaction * Beat out competitive offers * Growth need strength can inflate or deflate relationship between JDS and job satisfaction. *How much can the work itself effect job satisfaction?* -Fyre reported .50 correlation between Job Characteristics and job satisfaction. Also satisfaction with the work itself is the biggest predictor of outcomes. *How do you measure it?* - JDI facets & overall is popular (Smith). --0. Work itself --1. Pay --2. Promotion opportunities --3. Coworkers --4. Organization overall --5. Supervision * JDS (Hackman & Oldham). *Major models?* 0. Herzberg's motivators (responsibility and achievement) and hygienes (work conditions, pay). Note: Doesn't hold up quantitatively. 1. Lock 1976 defined the construct as both affective and cognitive. Match between what we want from the job and what we're getting from the job. 3. Lord & Foti 1984 - Social information processing theory holds that 2. Judge 2001 - Reciprocal relationship between job satisfaction and affect. We shouldn't measure those components separately. What cites do you need to know? * Locke (1976) - defined the concept. * motivators and hygienes (Herzberg) * Social information processing (Selancik and Pheffer) What are 3 recent studies? 1. Small group reflection in physicians: Boosted engagement and empowerment, but didn't touch stress or job satisfaction. 2. Mindfulness intervention reduced burnout. 3. Physical activity intervention did not improve job satisfaction. *Is a happy worker a productive worker?* Yes, if the job is complex and the worker's affect and cognition are linked [could imply metacognition as a trait]. According to a seminal meta analysis by Judge et. 2001, yes. .30 But there is such a huge standard deviation in validity coefficients that there are clearly lots of moderators going on, like job complexity, study quality, and which measure is used. Harrison et al: Job attitudes .59 with job performance Main cite for job attitudes Campbell 1963 *What is the consensus?* 1. The work environment is composed of different things. 2. Individual perceives something or combination of somethings about the work environment and compares it to some standards (expected outcomes, ratio of inputs, other people). 3. Some kind of mental calculation results in a difference value between perceptions and internal standards.

Strain (definition and practical tips)

*What is it?* Harmful and maladaptive responses to stressors. (Jax) *How do you get less of it?* (Chang & Johnson meta) -Emotional strain is easier to prevent than cure in the moment, due to fast cycle time of emotions. -Increase job resources for emotion regulation

Motivation

*What is it?* Latham & Pinder: A set of energetic forces within/outside that initiate behavior. 1. Content - Internal stuff. Needs (IM, SDT, Maslow), goals, motives. 2. Context - External. Culture, Job Characteristics. 3. Process - Goal Choice (TPB, VIE). Goal Striving (Self-Reg, Self-Efficacy), Goal setting (resource depletion models) *What is the opposite of it?* -Nueroscience: Schizophrenics feel motivating rewards with no action. Depressives don't feel any motivation. *How does it work cognitively* - Dopamine & testosterone - Self reg: Discrepency reduction (DeShon & Gillespie; Weiner) - Decision making - Attention - Goal is to direct energy towards the most effective ends at all time - Cost-benefit analysis - Interconnected goal hierarchies (DeShon & Gillespie) *Individual differences?* -Schizophrenia -Depression *How do you sell it?* -IM & EM negatively correlated -IM more morepowerful for work outcomes (performance, turnover, etc.) (Kuvaas) -EM unrelated or negatively related to many positive work outcomes *How do you get more of it?* -Goal setting interventions (Participatory goal setting vs. assigned) -Work redesign (Grant) -Expectency training (Pygmalian boot camp) -Self-regulation training -Goal-orientation training *In teams* -It's multi level -Attention alternates between individ and collective goals Cites -(Chen and Kanfer) - Team motivation ML How do you measure it? -Depends on what aspect of motivation: Work design questionnaire, IM measures, etc. *Cites* -Kanfer & Russ 100yrs -Deci & Ryan -Maslow -Fishbein & Azjen -Hackman & Oldham -Grant RJD -Bandura for self-regulation -Chen & Kanfer team motivation

Centering

*What is it?* When you have multi-level data, how you center has a big determining factor. Why? -There's group level variance when you have multi-level data, and you don't want to overlap variance. #Goals You're trying to change the X intercept. How: 0. Don't center: The expected Y value when x=0 --But sometimes that doesn't make sense (motivation = 0 isn't really a thing) 1. Grand mean center: xij - average X. You subtract the mean of x from all the x's. --Intercept becomes: The expected Y when individual has x value of the mean of X. --Mean becomes the intercept. 2. Group mean center: You subtract the group mean from each x in the group. (in MLT). --Intercept = Expected Y with an group mean level of X. --Helps you compare across groups. Debate: -When do you grand mean center and when do you group mean center? -Jeff: You GroupMeanCenter when you have a lev2 moderating variable on the relationship between two lev1 variables. -Grand mean center when you want to see the incremental impact of a lev2 var on a lev1 var that is also affected by lev1 var. So grandmeancenter controls for the lev1 var. -Team efficacy through self efficacy to performance = Grand mean center. -Separate model: Examining within and between separately, or when you're examining within-group and there's a main effect of the context. --You group mean center because you need to parse out the within-group variance from the between group variance. Cites: Hoffman & Gavin 98

Abusive supervision

*What is it?* You perceive your supervisor as being reliably hostile to you, but not in a physical way. *How does it work cognitively?* * Modeling and reciprocation. You fill somebody up with abuse, they have to release it somewhere. *What individual differences affect it?* -Authoritarianism (Tepper) *How do you sell caring about it?* 1. CWBs 2. Perceived organizational support as outcome/moderator 2. Interactional justice 3. Procedural justice 4. Lower affective commitment 5. Higher continuance committment How do you manipulate it? * Get a Low A, Low C person to confront the boss How do you measure it? * Tepper's (2000) measure is the gold standard * Warning: Outcomes are very context-sensitive *What cites do you need to know?* Tepper (2000) - Any conceptual stuff on abusive supervision. Plus measure. Mackey et al. - Meta. Outcomes and stuff.

OCBs

*What they are* * Discretionary behaviors that benefit organizations and their members (Organ 97) *Why they're great (Chang & Johnson meta)* - Help orgs build & retain top talent - Help employees build social capital - 25% of variance in company's finanical perf accounted for by OCBs (Podsakoof et al in Change & Johnson) *Things that crush them:* - Work strain (Chang, Rosen et al meta) *Ways to get more of them* -Maximize collective self-concept and individual self-concept integration with org. -Justice cilmate -Task engagement -Transformational leadership -Positive affect

Selection & narratives

1. Bobko et al: .30 correlation between biodata and job performance. 2. Shapes can predict, like redemption narratives predicting Schwartz values. 3. Hard to include them as an extra battery because they include ability, personality, etc.

Testing for publication bias

1. Funnel plot for things conslidated in the cone.

Test construction

1. Item generation: Content validity. Link to theory. 2. Survey: Sample who you want the scale to apply to. 3. Test structure: Unidimensional? 4. Reliability assessment. 5. Check construct validity.

Test bias

1. Non-cognitive predictors like personality/motivational/interests (Bradburn et al) 2. check for equivalence 3. Change recruiting strategy

Leadership development

1. Permanent, all-level leadership development (not just executives, not just socializing to peers) 2. Most leadership dev methods beneficial: Coaching Mentoring Networks Active on-job problem solving Job assignment

Longitudinal SEM

1. Use a recursive model in formulas 2. Give 1 latent variable per time point. --May need to use more variables to account for missing time points. 3. Latent curve model: Assumes that there is a trajectory equation.

Commitment

How does it work cognitively? 1. Cognitive - I think it's wise to stay here 2. Affective - I identify with this org 3. Continuance - I have to tuff it out. What individual differences affect it? * Old, uneducated, married women most committed. * Less related to turnover than low job satisfaction is, because it's more about being open to leave than necessarily wanting to leave. (Tett & Meyer) How do you sell caring about it to executives? * Large corr with turnover intentions. Turnover is a BFD because can pull other people down (Hom et al JAP 100 yrs) * More socially connected and rewarded within the organization. How do you manipulate it? * Collective self concept (affective) (Johnson & Chang) * Individual self concept (continuance, because personal invesment) * Value congruence * Job satisfaction (all facets) * LMX & coworker relationships * Abuse workers (creates continence commitment) * POS How do you measure it? * Meyer & Allen 3 component commitment scale What cites do you need to know? * Meyer & Allen - 3 types of commitment What are 3 recent studies? * Trust is a big antecedent, specifically interpersonal trust. Build interpersonal trust (Nyhan) * Half-assessed stress intervention actually reduced commitment

Interests

4 phase model: Hidi *What is it?* -The way you meet your goal of achieving your values. *Cognitive roots?* -4 phase model: Vocational vs. Situational interests --Situational: Really interested in talking about research. Varies broadly within person (e.g., interested in one class, not another). --Vocation: More stable, trait like, consisted. --4phase: People transition from fleeting situational interests to stable vocational interests. --Implies: Exposure to situations determine what interests you can gravitate towards. --Image: Galaxy breadcrumbs. -Multiply determined *Sell it* -VALIDITY = FIT with job --That means environ needs to go in model too. -Commitment & fit -Buffer against other things (e.g., quitting) --Interests are attractive. If you're interested in something besides your current job, it's going to threaten your committment. --Commitment, if you're interested in your current job OR if the job gives you time to work on your hobbies. -People tend to increase in social interests over time, probably because their work environment becomes more social as their career evolves (e.g., more management). Models? -RIASEC: Broad interests (more stable). Like Schwartz for interests. Uses mapped distance between hexegonal points. --But too broad. -SETPOINT/CABIN: Basic, specific interests (more flux) -General Occupational Themes (RIASEC), Basic (CABIN), Occupational (I want to be a teacher) 1. Holland's RIASEC model is dominant. 2. Hogan's conformity and sociability based on RIASEC 3. Tracey & Rounds (1996): prestige values 4. Interests predict career choice well Problems? -All intercorrelated --Interests don't differentiate from each other well in CFA. vs. Values: -Values are cross-situational -Values are what you think are important -Interests are how you realize those values Measure it? -Rank interests -Correlate items with people already in an occupation Future directions? - Incorporating technology in everything. - Maybe keep updating interests. Cite * RIASEC * SETPOINT (Su) is more modern

Selection

ASA model: Who introduced it? Schneider *What are some key individual differences?* Schmidt & Hunter 98 - Anytime you mention an individual difference as predictive of job performance. Schmidt & Hunter 2004 - Job complexity moderates relationship between G and job performance. Carroll (1993) - Developed modern structure of intelligence, with general factor, fluid/crystalized, retrieval. *What are the most effective selection tests for job performance according to Schmidt & Hunter 98 AFTER accounting for G?* 1. Integrity tests 2. Conscientiousness 3. Work sample *What are the most effective selection tests independently of G according to Schmidt & Hunter 98?* 1. Work sample predicts better than G 2. G = .5 3. Structured interviews 4. Job knowledge tests 5. Peer ratings 6. Conscientiousness (.2) 7. Neuroticism [not in S&H] Cite for personality and job performance -Barrick & Mount 2001 *What DOESN'T predict job performance well?* 0. Interests (according to S&H) 1. Age 2. Graphology *What are 2 wicked problems in selection?* 1. Unreliability of the criterion. Supervisor ratings are unreliable. 2. Range restriction. We're not going to hire randomly or hire all people. *How do you reduce adverse impact of a test? (4)* 1. Use non-cognitive predictors. They typically have less adverse impact, and still some predictive validity. 2. Focus on testing or controlling for motivational and educational factors that influence test performance. 3. Don't fix it. Prove job relevance of G. 4. Maybe: Get into subfacets of G.

Meta analysis

Add to the catapiller: 1. Add buffet: What concepts you're interested in. -Publication bias 2. Keyword nose: -Conference papers? -Different fields On moderators: -Any study design factors: Lab vs. Field, etc. Hunter & Schmidt bare-bones meta analysis: -Calculate the d for every study in the meta -Calculate the mean d across studies, weight by sample size. -Calculate the variance across studies, weight by sample size. -Adjust variance for sampling error. -You get an average d and an SD Ways to detect moderators: 1. If variance accounted for BY SAMPLING error 75% then probably no moderators. 2. Separate the studies by suspected moderator (e.g., lab vs. field), and if the d values move apart, then there's most likely a moderator, and their individual variability is lower than the overall variability, then probably a moderator. 3. Correlate moderator variables directly to effect sizes. Conceptual points: It's about averaging out sampling error. The more samples you have, the more you're averaging it out. -Moderators add systematic variance. Cohen's D: 2/5/8 s/m/large

Regression

Assumptions? -All assumptions are about the ERRORS and are theoretical. Regression will still work, but you shouldn't infer anything if any of these are true. 1. Errors are normally distributed 2. Errors are independently collected 3. Errors are uncorrelated between X & Y 4. Homogeneity of the errors (no heteroscadascity) 5. Model includes correct and relevant variables 6. Parameters enter the model in a linear fashion General model y = Beta * x + error Variables -Beta = .5 = for every 1 SD increase in age, you get .5SD inch in height. Types of models: - OLS - GLS - ML - Bayes (bootstrapped) Types of regression -Simple (cont/cont) -Multiple (multi predictors, any type of X you want) --Use it to figure out what IV's were important. -Multivariate: more than one Y -Hierarchical -Stepwise Hierarchical vs. stepwise -Hierarchical: variance gets assigned at each step -Stepwise: Like automatic hierarchical until the best model found. Computer does the work so you don't have to. Best practices -Strip outliers in both IV/DV -More participants than IVs, otherwise you overfit -IVs shouldn't be super correlated, else big standard error -Check for normality of residuals Multicolinearity bad = IVs are correlated, so when you control for each other they flatten their predictive power. -Inflates Type 2 errors -Inflates standard error -Leads to innaccurrate regression coefs *Logistic regression* -Categorical Y -Gets probability -Uses ML estimation -Predictors don't need to be normally distributed -Cannot produce negative probabilities -Best practices in logistic: --High ration of cases to variables --Multi -Probit logistic regression: Good for getting threshold values, but requires normally distributed X's.

Team Adaptation

Burke proposed an input-throughput-output model of team adaptive performance Cite? -Burke (2006)

Frameworks for understanding diversity?

Chao cultural mosiac -People influenced by a network of common and different cultural ties. D.A.G Demographic (physical) Geographic (natural/man-made temperature, urban/rural) Associative (groups you associate with; family, religion, profession, politics)

Climate vs. Culture

Culture: Shared values and assumptions about *WHY* ppl do what they do. Climate: A summary of *interconnected experiences* about what is rewarded, punished and supported. Commonalities: -Episodic memory and artifacts Differences: -Culture more complex and unconscious

Prejudice

How do you fix it? 1. Interpersonal contact 2. Exemplar counter-example 3. Increase cognitive resources 4. Boost mood 5. Give allies a voice 6. Laws 7. Self-labeling 8. Induce motivation to change

Team processes

How do you optimize it? -Emphasize accuracy of TMM as much as sharedness Cites? -Bell

What else increases power?

Increase treatment Decrease noise

Rick to court

Just Learn the foundations... -Multi-level modeling -SEM -Regression -Meta Focus on those... Focus on the current debates and hot things. -Look in ORM and see what people are talking about. -Lee's Biases in Meta analysis. What is the correct way to do meta-analytic SEM?

Hierarchical regression

Literally just entering your variables in a stepwise fashion and looking at the change. (regular regression is just checking to see if beat is sign, not change in beta) You add in variables one at a time like you did in multiple regression. The only difference is you actually GET a change in R value.

What are the threats to statistical conclusion validity?

Low power Range restriction Violated assumptions of statistical tests Fishing/error-rate problem (repeated tests without correction) Unreliability of measures Unreliability of treatment implementation Extraneous variance in experimental setting Innaccruate effect size estimation

Missing data

MCAR, MNAR, etc.

What are the threats to internal/external validity and how do you fix them?

Maturation - Control group Regression to the mean Selection - Random selection Selection X Maturation - Treatment/control groups are different. Solution: Random assignment. Mortality - Random assignment Instrumentation - Heart rate monitors were reading high or off in one group/session. Solution: Multiple measures. Testing - Same IQ test gets learned. History - 9/11 happens during a stress intervention.

Team leadership (short)

Quick points 1. Hackman & Wegman: Team leaders should set a clear direction, focus on hunting resources for team. 2. Marksmathieurzaccaro: Leader functions influence cog/motivation/affective process and facilities emergence of collective expertise. --Facilitate shared mental models. 3. Kozlowski & Chao: Change function as team matures. --High LMX = Sharedness of Climate 4. TFL in teams facilitate team-level motivation constructs like potency, but lots of moderators in team-level TFL that need to be nailed down. One tip: -Clear mission briefs = shared mental models

Training & Transfer

Sell it? -Investment in training related to firm profit growth (Ford & Prasad) *(Ford & Prasad) What is it? * 1. Generalization 2. Maintenance *Individual diffs?* -G & C -Mastery -Motivation -Pre/During/Post-training self-efficacy *Training design:* -Multiple learning strategies -Positive/negative examples -Inc error management -Space practice, progressive difficulty - Match technology to training goals - Personalize to learner *Work environment* -Prep leaders and peers to support transfer -Opportunities to apply skills immediately *Followup measures* -Follow _multiple_ trainees and find out what worked/didn't. * Koz & Bell for anything related to tech and training

IRT

Taylor: It's important to know the models. What is it? -ICC: Item Characteristic Curve. Relationship between latent trait and response. -Errors are uncorrelated 1PL model: -Assumes discrimnation param is constnant across all items and guessing is zero 2PL model: -No guessing 3PL model: -Accounts for people guessing: Somebody who based on their other reponses shouldn't get an item this difficult correct, but they do. -Guessing looks like flatting out (but above zero probability) on the left side of the ICC #goals -Get theta: Write items that assess different levels of. Discern your exact level on an underlying trait. -Maxmize item discrimination. How does it work? --Supposed to be able to look at a curve and tell where a and b and C are. -Discrimination param -Difficulty param -Pseudo-guessing param How do you do it: 1. Check item difficulty & discrimination 2. Look at ICC curves, hoping to see S curves at each ability level (y=probability, x=theta) 3. Look at TEST INFORMATION to see where your overall test is giving you the most information (at what theta). *Best practices* -Run 2PL first then 3PL. Simpler model is better if possible. -Use ANOVA to see if they're NOT significantly different. If not, KISS. -n=300 PER THETA -no reverse coding vs. CTT CTT -Item difficulty is 0-1, optimal at .5 -Difficulty determines variance -Discrimination: People in upper level - lower level. -Assumptoins are weak: scores should be normaly distrubuted, obs intdependent Advantages of CTT: -Most data sets can meet Disadvantages of CTT -Observed and true scores are dependent on the sample Nuance -Needs to be unidimensional ITT better than CTT at: -IRT item population params are invariant. -Thinks about your standing on a latent trait (vs. just here are some math questions) Misc: -Polytomous IRT = more than correct/incorrect. For personality tests instead of like SAT questions.

Validity

What are the types of validity and how to achieve each? 1. *Ecological validity:* It means that the study context and everything in it must be realistic (like, approximate the real world). One example: Set up mock office (highhouse) 2. Face validity 3. Construct 4. Content

LMX

What individual differences contribute to it? * Personality match How do you sell it to execs? * JOB SATISFACTION * Commitment * Role clarity * Turnover intentions (but not turnover) Who do you need to cite? * Outcomes: Day * Concept: Graen & Uhl-Bien 1995

Self-regulation

What is it and who invented it? How do you get more of it? * Goal framing * Goal content * Goal proximity * Teach emotion control strategies How does it work in teams? -DeShon & Koz -Individ vs. team goals, fixed number of footsteps (resources to allocate to either individ/team goal), handing out footsteps delimma (who should get them?), breadcrumbs as feedback -Discrepancy reduction a force: You focus on the goal that's farther away. Cites * Koz & Bell - DeShon & Kozlowski

Self regulation

What is it and who invented it? -The self's capacity for override altering it's own behaviors. (Baumeister & Vohs) Cognitive roots? -cybernetics/Control theory -self-monitoring -self-evaluation -self-reactions -self-efficacy Individual differences? -Mastery goal orientation How to sell it? -It is literally essential to the attainment of all goals. It is the mechanism by which goals are attained, perhaps especially difficulty ones. Who to get more of it? (biggest) 1. Well-defined standard/goal 2. Monitoring feedback loop 3. Energy/Self-reg strength 4. Motivation How to measure it? -Marshmellows eaten -Feedback seeking -Standard clarifying -Changes in effort -Self-efficacy *Cites?* -Wiener: Cybernetics -Carver & Scheier: Control theory -Baumeister & Vohs: Modern self reg -Latham & Locke: Goal commitment and self reg -Kanfer: Components in self regulation

Dynamics

What is it? 2 types of data: -Time series: 1 unit over time. Within-person. SEM for 10 days on same person. -Panel: Set of time series, not necessarily at same time point. Each has its own random effects. 10 people over 10 days, but different 10 days. *types of analysis* -Autoregressive or 'lagged' relationships are the base. --Can also be multi-variate (multiple X's and Y's). *types of stationarity* -Difference stationarity: if false, random walk -HLM is not enough: Describes changes in Y over time, but is descriptive not causal. Recursive rather than cyclical. X on Y and go. X affects y and it keeps moving forward. #Goals -Model CYCLES and use your knowledge of how the cycle works & vars in it to predict the future. -Ensure stationarity: a system that ultimate cycles around a homeostatic point. *When is it helpful?* -Data driven vs theory driven, no assumptions like in SEM -Captures cyclic relationship that people mean when they say "I study dynamics" but don't actually use dynamic modeling. -Easy to estimate using OLS *How to do it* 1. Select variables to include in theory 2. Check for 2 types of stationarity --Trend stationarity ---Trend good, random walk bad ---Detrend: Fit a trend line, then you remove the trend, leaving yourself with the random variance that's left. ---Gremlin: Changes in the mean violate stationarity --Difference stationarity: random walk bad 3. Select a lag length. --Use model fit information 4. Estimate model parameters using regression --So you're estimating the relationship via regession. So coeff, means, error. 5. Interpret. --If one X and on Y: Granger causality ---Granger-causality "this is useful in predicting the future". Or "T4 is useful in predicting T5" --Impulse response analysis: You shock the system by drastically changing one variable and showing that the other variables don't move. *Example* 1. Perf & self-efficacy 2. Dickey-Fuller test for stationarity --Trend stationarity, so vector autoregressive 3. R detects optimal lag length 4. OLS to get regression coefficients 5. Granger causality test. Performance granger-causes self-efficacy, but self-efficacy does not granger-cause performance. *Gremlins?* -Many time points required -Simultaneous relationships can be difficult to interpret -Requires trend stationarity. -Still working on panels. *How to do it?* Time series:

template

What is it? How does it work cognitively? What individual differences effect it? How do you sell caring about it to executives? How do you manipulate it? How do you measure it? What cites do you need to know? What are 3 recent studies?

Workplace hostility

What is it? * A continuum of prejudicial behaviors ranging from neglect to outright hate and mobbing. * Incivility (ambiguous) * Microagression * Harassment * Abusive supervision * Bullying * Mobbing How does it work cognitively? * Stereotypes * Out-grouping * Prejudice * Beliefs * Status What individual differences effect it? * Social dominance orientation * Genetic culture (terry cloth/china baby) * Anything that boosts status and low C How do you sell caring about it to executives? * Strain - Creates massive emotional issues (even PTSD in some cases) * Burnout * Turnover * All bad things How do you manipulate it? * Mood * Interpersonal contact * Readiness to change * Cognitive capacity * Labeling * Laws * Allies How do you make it worse? 1. Big power differential 2. High frequency 3. Unsupportive org climate How do you measure it? * Best to use multiple-methods (Naylor) * Harrassment - Sexual Experiences Questionnaire * Abusive super - Mackey measure What cites do you need to know? * Behave (Sapolsky) * Measurment & outcomes (Naylor)

Diversity

What is it? * Demographics (black/white) * Perspectives (rich, poor, skillsets) How does it work cognitively? * Monkeys have outgroups (Sapolsky) * People really good at categorizing (Keita) * 3d object: Beliefs and confirmation bias create different worlds * Identity What individual differences affect it? * Demographics * Tolerence * Survival pressures (Inglehart) * Openness to experience creates more tolerence * Scarcity *How do you sell caring about it to executives?* 0. Creativity 1. Defeats groupthink 2. Diversity bonus (decisions) (Page) 3. Stability 4. Reslience * Also creates fit and job satisfaction if your workforce is diverse. (Findler) How do you manipulate it? * Recruitment * Avoid adverse impact * Create open flow of ideas from bottom to the top * Cross-pollenation opportunities * Diversity training What doesn't work? * Diversity training can make What cites do you need to know? * Chao & Moon - Mosaic. People have multiple identies that all interplay in how they interact and form networks in the workplace. * Diversity Bonus (page)

Common method variance

What is it? * Noise in a relationship caused by using the same measure or setting to measure X and Y. * Perfect world vs. reality When is it useful to know? * Anytime you're trying to assess a relationship between an X and a Y. How do you do it well? * Multi-trait, multi-method matrix * Write items better * Separate X and Y * Surrogate * Common factor How do you sell its importance to executives? * Ackerman shorts story * Cote and Buckley: ** 40% of attitudinal measures were just method variance. ** 20% of job perf was method variance Who do you need to praise when you use it? Podsakoff - Wrote the big paper on it with solutions. Not MTMMM though.

Leadership effectiveness

What is it? * Typically: A leader's impact on the organization's bottom line. * Better: By the performance of people he leads What individual differences affect it? * Motivation to lead predicts KSAs * Agreeableness (LEffect) * Openness (LEffect) * Values (TFL) * Beliefs (transactional vs. tfl) -Mid-level narcissism (Hogan) How do you manipulate it? (biggest first) How do you measure it? -Best: Performance of the group they lead. Who do you need to cite?

Work engagement

What is it? * Vigor * Dedication * Absorption Cognition: * Evolved tendency to broaden & build * Dopamine/reward Individual differences: * Core self evaluations Sell it * Burnout, though separate poles * Similar relationships to ORG performance (.3) (Shmidt meta analysis) * More related to task perf than JS * OCBs How do you manipulate it? (most powerful first) * Job resources intevention (highest max effect, but -.04 to .84) * Value congruence * Health promotion (most reliable without CI = 0) * Group and individual interventions work How do you measure it? * Empty room study * Utrecht work engagemet study Who do you need to cite? * Kahn: Definition * Shmidt: Related to performance * Lepine et al: Antecedents * Knight et al: Interventions

Leadership emergence

What is it? * Who will be promoted? * Also: Who will spontaneously emerge as leader in leaderless groups? How does it work cognitively? * Stereotypes * Availability heuristic * Familiarity & liking What individual differences affect it? (Judge, Bono et al) Men who are high on all the positive Big 5, are authoritarian, and have performed well in their current role. How do you sell it to executives? * You can get more effective leaders by checking your stereotypes at the door. How do you manipulate it? How do you measure it? * Promotion, usually Cites you need to know. * Judge, Bono et al - Personality and leader emergence

Systems Theory

What is it? -The notion that everything is connected to everything. -A way of seeing organizations. *How does it change the way you look at orgs?* -It takes a connection between two parts to make a system. -Closed systems don't take anything in from the outside. Trend towards disorganization. -Open systems exchange a lot of the outside. Trend towards organization. How does it affect research? -To be fully realistic, everything we ever do should be multi-level and network based. -Subsystems thinking vs. holism: Researchers can only focus on one subsystem (couple gears) at a time (like the drunk under the lightpost), but systems are not just the sum of parts.

Multi-level theory

What is it? - Organizational data is inherently nested - Lewin: Interaction between P-E - Emergent effects (compilation & composition) - MLT helps you conceptualize and study What do you need to know? - Level of origin: - Forms of composition: 1. Additive model (just average them) 2. Direct consensus (amount of agreement) 3. Referent shift 4. Dispersion - Variance or agreement is a construct on its own (diversity, power distance) 5. Process - Analogous process. Team memory works like individual memory. Same structure and function. How do you do it? - MLT Womabat X. OPERATIONALIZE YOUR CONSTRUCTS What type of constructs? --Global, shared, configural --Global cannot exist at any lower level than it does X. PICK YOU MODEL a. Single-level (team efficacy on team performance) b. Cross-level (direct effects: lev2=>lev1; cross-level moderator; frog pond) c. Homologous multi-level (team efficacy affects performance just like individ efficacy affects perf) X. PICK YO SAMPLE --What people do you need to involve to operationalize your constructs? X. PICK YO MEASURES --Referent shift? X. CHOOSE YOUR ANALYSIS X. JUSTIFY YO AGGREGATION X. ICC1: See if the level 1 variable is shared enough to be aggregated into a level 2 variable. Measure of non-independence. If > 0, there's within-group variance that needs to be accounted for. Level 2 variance (between group) / total variance (between + within). If it passes, then... X. ICC2: Gives you empirical support for aggregration. Tells you if the means at level 2 representative of the individual scores. --Want it to be large. X. RWG: Agreement at a higher level (can ratings from any one individual be swapped out without affecting the group rating). How high is agreement on a given variable within a given unit? Others: -- WABA: Within and between variance. Another way to justify aggregation. -- eta squared: Another way to justify aggregation. --HLM How do you do it over time?

Self-efficacy

What is it? - The degree of your feelings about your ability to accomplish your goals. (Bandura) *Cognitively?* -could be just a pure association with the notion of "you performing" -Past performance more correlated with SE than SE is with future performance. *Indivivdual diffs* -Core Self Evaluations Selling it -Strong relationship with work performance -Especially true with SIMPLE jobs. Less true for complex jobs. *How do you get more of it?* 1. Increase signal (of progress) -Self monitoring, feedback, and social support (e.g., Fitbit social) 2. Decrease noise -Role and task clarity (clearly explain the assignment) -Make resources for accomplishing it known -Increasing processing resources --Cut distractors --Stress management -Teach workers how to learn/train themselves to overcome big obstacles. -Help workers frame their abilities as malleable. Measuring it Cites -Bandura -Luthans for outcomes

Longitudinal

What is it? -3+ time points How does it help you? -Establish causality well -Study CHANGE in all variables #goals: -Fully capture change, not just approximate it with lines How do you do it? -Nonlinear all the things, to approximate well X. Figure out the shape of your curve, which is how you're going to scale time (polynomial or orthagonal?) Best practices: X. Clearly specify the change you expect to see in your h's/ Gremlins: -Threats to internal validity --Word of mouth --Attrition -Threats to external validity --timing --Context --Did you capture the action?

HLM

What is it? -A general term for a broader class of models when you have DVs at the lowest level (climate predicts individual perf) -Uses random coefficient modeling, which is letting the slopes and intercepts vary freely within-group, and then using them to inform the next level of analysis. What does it do? -Step 1: Does a regression for each person (performance~personality over time) -Step 2: Uses the personal coefficients and intercepts as the Y variables when we do between-people values New Y~Climate HLM vs. OLS: -OLS does the same thing but uses average coefficients across groups, so doesn't account for individual/unit error terms. -HLM does not violate assumptions of independence; DOES account for error terms. So doesn't inflate the error. When to use: -You know this from Kelsey. -Frog-pond/cross-level CARMA video super helpful -Make an account because we have an MSU membership

Climate

What is it? -A summary of interconnected experiences about what is rewarded, punished, supported, and expected. Cognitive roots? -Gestalt -Attention -Punishments & rewards Individual diffs? -Safety cilmate -Justice climate Sell it? -Team outcomes stronger when climate is highly shared. -Justice climate gets you OCBs and LMX Manipulate it? -Leader attention (Zohar) Measure it? -Sharedness/agreement gets you climate strnegnth Cites? -Most things: Schneider -Leadership drives climate (Kozlowski & Doherty) -Safety culture (zohar) Future directions? -Climate as the 'crucible' for multi-level theory. (Koz & Klein) -More leadership (Kozlowski) -Integration of climate and culture

Transformational leadership

What is it? -Burns: Motivates followers to achieve performance beyond expectations by changing attitudes, values, and beliefs. -Bass: Subdimensions (not a lot of suport) --Charisma (idealized influence) --Inspiration --Intellectual stimulation --Personalized consideration Cognitive roots? - Need fulfillment - Values - Beliefs - Pygmalion effect *Individual diffs?* -Schwartz Values -Openness -Agreeablness -Conscientiousness UNRELATED *Sell it?* -task performance and OCBs -organizational performance *Get more of it?* -Transformational leadership training can work (Brown) -Self-regulation design -Idea/referent: Who's the best leader you've ever seen -Planning/Feedback: What behaviors will you enact? -Application *Measure it?* -MLQ *Challenges?* -Active/passive dimensions fit just as well as the 5-dimensions proposed by Bass *Future directions?* -Figure out actual dimensional structure *Cites?* -Creator: Burns 78 -Dimensions: Rafferty et al -Outcomes: Wang et al

Personality

What is it? -CNye: Preferences for certain types of behavior Future? -Compound traits --e.g., Integrity = ACEs --Predicts broader range of outcomes --More efficient: 10 items vs. 30 by cherry-picking from lower order traits. *How do sell it?* -Your environment has a personality, and will activate certain traits ([trait activation]) --Matching personality to job boosts validity (Hogan) *Personality also predicts* Job performance Training performance OCB's CWB's Leadership Performance *What are some advantages to using personality to predict things?* 1. Less adverse impact 2. Can predict a wide range of work outcomes *Best practices* -Facets predict job perf better than general factors (and help clarify overall traits). (Roberts) -Figure out what traits are most relevant for this job, create a compound of that, and see how well it predicts. -Only if weak situational strength Models: -Big 1: probably just measurement (Rushton) -Hogan's Big 2: Getting along, getting ahead -DeYoung & Peterson Big 2: Stability & Plasticity -Big 5 (NCA | EO) -HEXACO (Ashton) -Facets (DeYoung & Peterson). Less agreement. -Barrick & Mount Theory of Purposeful Work Behavior - Not a lot of evidence. --Trait activation theory works though (certain environs enact different environments) *Things that don't work:* -Non linear relationships with personality rarely work Cites to drop: -(most of this is from Connelly, B. S., Ones, D. S., & Hülsheger, U. R. (2017). Personality in industrial, work and organizational psychology: Theory, measurement and application in D2L) -[Judge et al. is a review of the hierarchy of personality. Just cite that.]

Faking

What is it? -Job applicants deceptively answering self-report batteries to appear more desirable. Cognitive roots? -Self schemas -External locus of control Individual diffs? - Performance orientation - G: Ability to fake - Motivation to fake Sell it? - FFM doesn't even fit in job appliants; the data can be that skud. - Measurement variance: Can shift means & structure in constructs - 30-50% of applicants fake their answers, and in doing so can actually CHANGE THEIR RANK ORDER How do you measure it? -Intention to fake -Social desirability scales -Uncommon virtues -Applicant reaction times How do you fix it? (Zicker & Drasgow) 1. Social desirability scales and impression management scales only modestly effective. 2. Forced-choice tests are more resistant to faking, but can only be used within-person because scores are ipsative. 3. Uncommon virtues scale 4. Speed up the scales. Nuance? -You can't say faking and social desirability are the same thing, at least not as SDesire is currently measured. Cites Neal & AMry - FFM doesn't fit job applicants. Griffith - Yep applicants fake like crazy Fine - Latency

Team Inputs/Composition

What is it? -KSAOs -Surface: Black box (age, race, etc) -Deep-level: Personality, etc. Cognitive mechanisms? -Homophilly & ASA *Individual diffs?* -Average G, moreso in complex jobs. -It's not individual personality/values so much as min/max/average -Minimum agreeableness -Mean C, O, Collectivisim -Preference for teamwork. Sell it? -Functional diversity of TMT = bank innovation, controlling for team/org size (Bantel) -Agreeableness and C more predictive for team perf even than individual perf -Higher GMA results in more accurate mental models. Ideal team composition? -No ******* threshold (min A) if interdep -Avg C -Max avg O, with one high O being good -Avg ES -Max collectivism -Max preference for team work -Maximize functional diversity -Diverse functional background -Some team tenure -Faultline bridge characteristics Measure it? -Min/max or average personality, etc. Cites/History lesson? -Outcomes of composition: Bell Nuance? -Big bucket problem: It seems like how you operationalize things like "team composition" and "team process" can differ wildly. -Bell: May not find personality effects in lab; only in field -Piece-by-piece problem: A Future directions? -Lots of moderators to be discovered

Difference scores

What is it? -Measure of congruence/overlap between two constructs. -Usually used as a predictor of some outcome *When is it useful?* -Any time you're measuring *fit* or *expectation vs. reality*. *How do you do it* 1. Subtract one variable from another, then you look at that diff on the sampling distribution to see how significant the size of it is. *Gremlins* -It includes variance and error from both the variables you differenced, which can be good or bad. -You don't where from the actual prediction comes from (which variable). Best practices: -Test the relationship you want to look at, not the diff itself. Use org's values as a moderator on a polinomial regression of individ's values on satisfaction. -Use polynomial regression to address some of the issues with difference scores. -Or use response surface methodology, because polynomial regression is hard to interpret.

Difference scores

What is it? -Measure of congruence/overlap between two constructs. -Usually used as a predictor of some outcome *When is it useful?* -Any time you're measuring *fit* or *expectation vs. reality*. *How do you do it* 1. Subtract one variable from another, then you look at that diff on the sampling distribution to see how significant the size of it is. *Gremlins* -It includes variance and error from both the variables you differenced, which can be good or bad. -You don't where from the actual prediction comes from (which variable). Best practices: -Use polynomial regression to address some of the issues with difference scores. -Or use response surface methodology, because polynomial regression is hard to interpret.

Affect

What is it? -PANAS (Watson & Clarke) -Stressful events -Affective events -Core emotion episodes -Emotional contagion -Mood: Broad, persistent, untargeted -Emotion: Short, targeted, brief. *Cognitive roots?* (Brief & Weiss) Affect: BIS and Behavior Facilitation System Emotion: Memory networks called up for certain emotions. Individual differences? -Nueroticism -Extraversion - PA - *Sell it?* -Negative affect and performance (Merlo) -Positive affect and creativity (construal level) -Positive affect and broadening of connections -Neuroticism/Negative affectivity affects structure of work networks. -OCBs and Turnover -Influences job satisfaction -Personality determines mood but not emotion. *More outcomes:* 1. Negative emotion: Basically, if you feel a negative emotion targeted towards work, you will want to avoid work. So withdrawl, turnover, absenteeism 2. Positive emotion. You broaden & build. More OCBs *Even more outcomes* Judgement Creative problem solving Helping behaviors Performance Ilies et al: Within-person variance in mood explained 26% of variance in job satisfaction within-person. Get more of it? -De-stress workplace -Hire less neurotic people (or provide emotional regulation resources) -Hire more extraverts Measure it? -PANAS Future directions? -Brain -ESM methods to model it over time -Person-centric -MLT Cites? -Brief & Weiss: Affect in the workplace (general cite) -Weiss: Affective events -Merlo: Affect & perf -PANAS = Watson & Clark In what order is an emotion formed by an event? 1. Event 2. Primary appraisal - Is this relevant to me? 3. Secondary appraisal -meaning. 4. Emotion

SEM

What is it? -Structual equoation modeling When is it useful? -It's all about the latent variables -Complex and causal relatoinships between variables, espeically latent variables How to efa: 1. Scree plot, look for elbows 2. Eginvalue: How much an item contributes to that specific factor. If > 0, then you usually retain that factor. Disadvantages to EFA: -Can't test specfic hypotheses -More art than a science -All or nothing correlated or uncorrelated in a model -Measurement error can't be corrected Parceling: -Good: Finding factors esp in non-normal data -Bad: Creates problems for measurement equivalance SEM/CFA models: -Bifactor model -Higher-order model -Formative --Implications: depends on what factors you choose, reliability and convergent/discriminant validity stop mattering. What do you even compare it to? --Completely meaningless and useless without outcomes. -Reflective. *Path model* -Series of regressions to see if one variable predicts another. -Gremlin: Make the assumption that observed variables are perfect and there's no measurement error. -After you change #goal of measurement model -Do our items create a nice latent factor? -Reliability can kill this *Gremlins* -Lower the reliability the more error -How to address measurement error 1. Corrected correlation matrix --Get better measure for the paths, but your standard errors will be off 2. Take out variables with low reliability --[Bad: Omitted variable problem. Makes you a ****ing liar.] 3. Constrain the model to account for measurement error. 4. Best practice: Use CFA to model measurement err. *Things you need to know:* Types of rotations: -Orthogaonal: All vectors remain at right angles; not allowed to correlate. --Verimax --Quartimax --Equimax --Oblique -Communality: Amount of variance explained by the factor Problems in SEM: -Bad: Linear dependence in matrix (singularity) -Good: full rank matrix Best practices: -

Validation

What is it? -Validate tests or selection tools How to do it: -Concurrent: You give people with high & low job performance your selection test. -Predictive: You give test to people who are not in the job yet, then check their performance in six months. X. Figure out which predictors predict best. -Hierarchical or stepwise regression probably. X. Create your final composite X. Answer these questions: -What is the predictive validity of your composite? -Still hold if you correct for unreliability in the criteria? -Adverse impact? X. Check for range restriction -Correct for direct range restriction: Create smaller version of your dataset, like just those above the mean on verbal ability. -Correct for indirect range restriction: Select above the mean only like education and see.

Values

What is it? -What you think is important, broadly (Kanfer) -A stable individual difference representing what people want to do or want to have in their lives. What are the two main values frameworks? 1. Schwartz 2. Rounds et al. work values What are the two main outcomes of values? (Dawis & Lofquist Theory of work adjustment) 1. Person-Organization Fit 2. Job satisfaction 3. Turnover How do you measure it? 1. All values are desirable, so you have to do them as a rank-order like the Schwartz survey does. 2. Hard to measure distance between a person and organization's values because difference scores are ambiguous and unreliable. (Edwards, 1993) Versus interests -Interests are how you reconcile your values with the real world. -In choosing jobs, your values kind of define the space, your interests make the choice. -Interests can BE a value.

Simple slopes

What is it? -When you add a moderator, you know it's significant, but you don't know where it's significant or by what amount it has an effect. You just have the slope change, but you don't know what the relationship really looks like. -Is there a sig difference in two lines in all places, or just in some places? -Simple slopes uses regression coef, your means, and your SDs to calculate where and how much the slopes differ. How to do it: -Just say "do simple slopes, SPSS". --Computer USES plus/minus 1SD to give you simple slopes (lines), and they will be significant or not AT plus/minus 1SD. ---Uses beta, mean, and sd. ---Computer says "at -1SD they are sig, at +1SD they are not." May or may not draw the lines for you.

Moderated Mediation

What is it? -You have a moderator squirrel (with shrink/enlarge ray) AND a mediator/human centipede squirrel -And maybe a direct stray from X to Y -Holland: But you need to specify where the moderator is coming in. Before or after the mediator?v Complexities: -Multiple mediators (stacked squirrels) -Time-bound mediators (cintapede-style mediator squirels) -Moderators that affect any or all links Best practices -Specify which link your moderators affect, and where your mediators appear in the causal chain. -Use experiment design or be able to clearly justify direction and causality. -Use measures that are low in measurement error because the more squirrels the more flies (error) -If you have a direct effect, you better be able to justify it. -If you're arguing full mediation, show that your direct path is nothing. -Report path coeffs for all levels of your moderator, or at least 2. *Special case: Moderated mediation* -You have to test indirect/direct paths if it moderates XM, but not if it moderates MY. --AKA if it moderates A but not B

Polynomial regression

What is it? -You're using X as its own moderator. -Level of Y is determined by X AND this effect changes with the level of X. When is it helpful? -When a straight line can't touch all of your data points. -Can be used as an alternative to difference scores. How do you do it? -Hierarchical regression 1. Enter linear term (X) first 2. Then inter an X*X interaction term (X * itself) and see if it predicts better. 3. With a moderator: Enter linear stuff first (X, W), then interactions (W*X, X*X, X*X*W).

Transactional leadership

What is it? Burns 78: exchange relationship between leaders and followers such that followers receive wages or prestige for complying with a leader's wishes. Transactional leadership encompasses contingent reward and management-by-exception

Measurement error

What is it? It's the difference between the true value and the measured value. It's random OR systematic variance in your measure that's not due to variance in the construct. What causes it? 1. Unexpected stability in construct 2. Transient states 3. Poorly written items. 4. Experimenter error 5. Situational factors What does it do? Typically, attenuates your correlations and your regression line. But can also inflate them. How do you fix it? 1. Multiple measures of the same construct. 2. Survey Quality Procedure helps you estimate measurement error so you can 3. SEM correct for it: "1 minus alpha reliability times variance of that scale's data". Then you pull it out as a latent var in SEM. 4. Go over responses with users to see if their responses are what they really meant. What cites do you need to know? Schumacker 2010 - How to calculate measurement error.

Workplace bullying

What is it? Repeated disturbance that blocks people from goals? How do you fix it? 1. Treat it as an org problem 2. Low status person may confront 3. Maybe: Confront v. Org support v. social support 4. Doesn't work: mediation because bully not interested in changing. How do you sell caring about it to executives? 1. Strain (because blocked goals) 2. Burnout 3. Absenteeism 4. Probably: Lower meaningfulness because of sysiphys.

Mediation

What is mediation? What are the ways to test mediation and what are the positives and negatives? *Barron & Kenny* -Test XY -Test XMY -If AC is not sig after adding B, then you have mediation Step 1: A: What is the direct of X on the mediator? If the mediator is going to explain something, then it has to have SOME relationship with the mediator. Step 2: C: What is the efect of X on the outcome variable? There should ultimately be an effect. Step 3: B->C If you control for the mediator, there should be an effect left over Step 4: The full model. You're looking for a change in C (called c-prime/C'). If C' is equal to zero, then you have full mediation. If you test the full model by controlling for M, and you have nothing left over, then you have full mediation. *Sobel* -Better than Barron & Kenny but also bad -Quantifies the indirect effect *Bootstrapping* 1. You automatically take certain individuals, and then you create new samples. You do that 1000 times. So you sample 500 random people from your sample 1000 times, and that gives you 1000 data sets, then you calculate the indirect effect (a*b). You then have 1000 estimates of a*b. 2. You can order those estimates from lowest to highest, and then use a confidence interval. If you want a 95% confidence interval, you take the 2.5% at the top and 2.5% at the bottom. Then you can say "95% of your data will fall between this range"

Team

What is thing? Cognitive roots? -Macrocognition How does it work? - Team life cycle 1. compositoin 2. forming (socializing) 3. processes 4. leadership & motivation 5. Continuance and decline -context, workflow, levels, and time Individual diffs that can affect? Sell its importance to executives? What can go wrong? How do you make it go right? - Train adaptability and learning itself -- - Error management training (crashed southwest flight) Who do you need to cite? * Kozlowski & Bell - Team formation / life cycle * Kozlowski and Ilgen

Harrassment

What makes it worse? How do you fix it? 1. Confront 2. Social support 3. Org advocacy

Meta-analytic SEM

When is it useful to know? -You have a bunch of studies on the same relationships and want to see their overall relationships. -Test hypothesized models with existing data How do you do it? 1. Correlation matrix bingo: Pool the variance/covariance matricies from each study together. Like, square-by-square if neceessary. 2. Use pooled matricies to fit the SEM model. *Gremlins* -Most gremlins in the pooling step. After pooling it's mostly a normal SEM. -Pooling problems: 1. Varying sample size problem: You're building a covariance matrix from different studies with different sample-sizes, and most researchers use the mean sample size, even though they vary. ---Regular SEM has one sample mean, so you're swapping in one overall mean for that. ---Can lead to over/under estimate relationship 2. Really want a covariance matrix for SEM, and meta-sem treats a correlation matrix as a covariance matrix, which many argue can throw off SE. *Best practice* -Use two-stage SEM (TSSEM) approach that uses Weighted Least Square estimation instead of ML, which addresses the SE concerns of using a correlation matrix as a covariance matrix.

Interactions/Moderation

X variable, Z variable (moderator). If Z is significant, then there's evidence of a moderation. last comps did ask about moderation in hierarchical moderation: -Multiply X & Moderator variable to create -Create the moderator variable (interaction term) by multiplying conscientiousness and cognitive ability together (so X*Z = interaction term). 2. Run a regression including the X, Z, X*Z. In hierarchical regression: - X and Z go in step one - X*Z go in step 2 If you add in your interaction term and the change in R is significant, then your interaction is significant. 3. Plot +/- standard deviation to see the interaction. Some people recommend plotting a 10% and a 90% score to capture more of the variance. 4. Formal test: Simple slopes test to tell you if there are significant differences in the points at the end. -At the value of SD (+/- one SD), then value is blank (and is that difference signifance or insig). -At each point in the line, are those two points siginifanctly different?

Work health cite

Zohar?

Classical test theory

observed score = true score + measurement error