Experimental Psychology Final Exam Review
intro to autism lab and research (intro/hypotheses)
(1) anxiety and autism: (a) is anxiety a core symptom or something that triggers core/makes aspects more severe (b) consequence of autism or something worsens symptoms (c) core: social communication: experience stress in social situation (autism -> anxiety); repetitive behavior: help calm down (disrupt sameness and disrupt autism core) (d) repetitive behavior in res to anxiety; sameness/repetitive behavior -> autism -> vul. to phobia/fear (e) Jenkinson et al 2020 (meta-analysis/review): 1/2 comorbid disorder make core symptoms worse/social comm; Kerns 2015 comorbid (social anxiety, phobias) + Kerns 2020 (2) wood et al 2020: autism vul to maladaptive + other study (3) use anxiety test to see how change subject's behavior; anxiety situation: core/repetitive behaviors worse (make autism symptoms worse) or co-expressed
anxiety test example
(1) anxiety test: (a) two dimensions/constructs: state (how feel right now) and trait anxiety (how generally feel, stable, personality) (b) some items had to be reversed: scores were flipped and counterbalanced, improving reliability (if I am pleasant = 4, put down 1) (2) score meaning: (a) need to compare to average (range is 20 to 80) (b) age: average varies based on age (c) compare to age-relevant norms and then interpret (d) after 2019/pandemic: norms are different, can't interpret based on these norms anymore (e) pre-2019 norms: low anxiety in college 33-36, low anxiety not in college 20-28, 60+ scores for 60+ age low anxiety (f) since pandemic: norms, state and trait anxiety levels changed (g) interpretation more questionable now (3) overall: (a) age and what doing at age matter (b) ex. higher anxiety in college students than those not in college b/c more deadlines (c) validate tests against other things like age, profession (compare to the right norm group)
qualitative: archival/interview/case study
(1) archival: (a) death records high for Parkinson's; in death records error tho b/c many factors go into it (ex. heart attacks from Parkinson's; COVID worsened breathing) (b) looked at regional rates, highest in southeast agriculture district b/c chemicals (c) points us where to look; can lead to correlational or quasi; shows us a problem in an area and then ask specific questions with quasi/corr and then do experimental (d) foundational gets us to more quantitative/exp. (2) interviews: what does thanksgiving mean to you? find themes for analysis (3) case studies: (a) 1 person in go; not much stat person; 1 person with rare condition (b) rare brain virus 4 cases/yr (c) 1 person in exp; control = as many normals as possible; low power (d) ex. Jarvis 1 heart; quasi-exp: control some level of heart disease (e) can set up case as corr/quasi/exp (d) usually rare/expensive condition
violating assumptions
(1) assumptions for all stats: (a) distribution must be normal (mean = median = mode; approx or exactly equal) (assumption of normality) (b) Pearson's R need 20 subjects at least for it to work (c) not normal: gp is dif even before start (2) ANOVA: (a) normal distribution, homogeneity of variance, sample size, same pop variances (b) robust: can handle if violate any ONLY IF it's a normal distribution (3) nonparametric tests to use when violate: (a) Mann-Whiteney-U-test for independent samples t-test (b) Willcoxon for repeated measures/correlated samples t-test (c) spearmen's rho for Pearson's R (d) wilcoxon-wilcox for 1-way ANOVA + post-hoc (not repeated measures ANOVA) (4) SZ example: (a) normal clustered on one end; more extreme with paranoia/catatonic on other end (b) can exclude extremes and keep general (smaller sample + less generalizable, but can still analyze) (c) later on exclude normals/generals and keep extreme (5) applies to all stats
experimental research 1/2
(1) between subjects designs: (a) randomization to all levels of IV; don't reuse subjects/all dif ppl in gps (b) any # of IV (c) no repeating or wn subjects factors (c) compare bet two gps: gp1/trmt 1 and gp2/trmt 2 (d) two separate gps (e) 1 DV with only 2 gps = independent samples t-test (f) 1-way ANOVA: bet gps ANOVA; for same experiment but ONLY two-tailed; t-test can be 1 or 2 tailed (g) t^2 = F; if just two-gps don't need post-hoc for ANOVA (2) 3 x 3 ANOVA: (a) more than 2 gps (no t-test) + need post-hoc (b) type of academic college (IV): science/engineering, liberal arts, nursing/health sciences (c) amount of laptop use (IV): none, HW, everything (d) analysis: main and interaction effects and post-hoc; ME of college/use and A x B interaction (college and use) (e) do simple test of ME first then post-hoc ex. Turkey (f) interpretation: can't RA to type of college to quasi-IV (likelihood); for research design look at how made gps and interpretation
correlational and quasi-experimental research ch.7 (in class)
(1) both only talk abt relationships; correlational = relationships bet DV; quasi-E = preexisting gps (2) correlational tells us: (a) do relationships exist, strength of relationships, prediction, analysis of naturally occurring behavior on a large scale, simultaneously analyze a vast number of variables (multiple regression and factor analysis), study relationships when it is not possible to ethically manipulate variables of subjects (3) correlational if see if relationship exists in the first place (4) example: (a) screen violence and actual violence (b) population: if use extremes (criminal record/juvenile delinquents) ceiling effect + confirmation bias (5) problem of directionality, third variable problem (confounding variables), cause and effect can't be determined
Reliability
(1) consistency, same results when taken multiple times (2) test-retest reliability: (a) measure with correlation coefficient bet two measures (b) take test twice, scores should be similar/consistent; 90/95% (c) but know answers now: so practice effect and know what's on it; assume: learned (d) limited by learning over time (3) split-half reliability: (a) same question written a little differently in second half (b) split test into two halves and see scores on first half vs. second half (c) similar content, ask different in 2nd half (d) ex. multiple choice: what is IV/DV; short answer: identify IV/DV in research design (e) if right multiple choice, right short answer; if wrong MC/wrong short answer
correlation cont. + ch.8 in class start
(1) correlation: (a) problem of directionality - which V influences the other; 3rd variable/confounding variable problem - another variable influencing both (b) CBD use predicts gender (males use it more), b/c of larger effect in males, more risky behavior, women exp. anxiety? (c) ppl w/o psych disorders attitude of drugs = aversive; preexisting psych disorder = 3rd V (d) Freud self-medicating and treated ppl with cocaine for depression (attitude = 3rd V b/c it worked for him) (e) 3rd V problem in all designs, but esp. in correlational b/c no controls like in exp. (f) no cause and effect can be determined (2) qualitative research: (a) emphasizes that the whole of human experience cannot be adequately represented through quantification (b) the emphasis is on the whole of the experience as a means to understanding behavior (c) file drawer research = analyze archival research/archives (d) areas where possible research design (maybe not best): archival studies (nonparticipant studies) = de-identified data (ex. state death records), electronic medical records (HIPPA regulations), letters and manuscripts (3) physical traces: (a) sweat for drug test; traces left of ppl in areas and identify it; nonparticipant studies (b) nonparticipant, don't need informed consent (c) your view influences how you see things; what did we find/any themes (d) independent auditor: not in process, determines what it is, look at objectively, reflectivity: researchers reflect on their experience as part of the research process (e) b/c own confirmation bias, need independent auditors (4) interviews/focus groups/case studies (participant studies): need informed consent
ch.7: correlational
(1) correlational: (a) relationships among variables (b) how are changes in 1 V associated w/ changes in another (c) categories or quantities (d) quantitative: can say something about direction/value: positive (both up/down), negative (opposite) (e) no manipulation (2) can tell us: if relationship exists; naturally occurring + no intervention; can measure large number of variables; if can't manipulate ethically/practically, can assess w/ correlational (3) non experimental research = no manipulation and no RA (4) time-series analysis: analysis of data points collected repeatedly over time; longitudinal design: same ppl followed over time (5) cohort-sequential design: combines cross-sectional and longitudinal: selected at dif. points and followed over time (6) cross-sectional design: dif. populations measured at same time
Ch.5 intro + scales
(1) criteria to evaluate measure quality: (a) validity = measures what it says it measures (b) reliability = produces same results over multiple administrations; all items assess construct (c) internal consistency/Cronbach's alpha: how each item in scale correlate w/ e/o (0 to 1) (d) length of instrument/item difficulty (e) cost/copyright/training level (2) construct: idea/theory whose properties are inferred from a measurement and are not directly observable (intelligence/anxiety) (3) use scales to quantify the abstract; will never have complete overlap of theory and measure, but want max overlap (4) constructs harder to measure the more abstract they are; use operational definition: define V by operations/processes used to measure/quantify it (safe driving by number of cars doing full stops at stop signs)
ch.7: cross sectional, cohort sequential, multiple methods
(1) cross-sectional: save time/money, but cohort effects possible (age vs. technology?) (2) cohort-sequential: both cross-sectional + longitudinal, measure generational effects (3) advantages of using multiple methods: designs complement each other
inter-scorer reliability
(1) done in lab 2: autism/anxiety in mice with dependent variables (2) use b/c practical: do less work in future, can have multiple scorers (3) consistently measures variables -> better accuracy and generalizability (4) control stuff: measure what say measure; better internal validity
ch.7: correlational drawbacks + quasi-experimental + stats
(1) drawbacks: (a) directionality: don't know which is cause and which is effect (A cause B or B cause A, C cause both); don't know which direction it flows in (b) third variable problem/confounding variables: impacts causality, impacts IV and DV relationship (C cause A and B?) (c) can't determine cause and effect (2) quasi-experimental: (a) dif bet preexisting groups (b) no intervention/manipulation; no cause and effect determined (c) can have true IV and quasi-IV in one study (if quasi-IV only talk abt associations NO C and E) (3) stats: (a) correlation analysis: relationship bet 2 V (interval scale) (b) regression analysis (estimate ability of V to predict an outcome) (c) both: sample as a whole (d) Pearson's r = stat measure of correlation bet 2 V (4) regression analysis -> predictor (V predicts outcome); criterion (outcome measure of interest); standardized regression weight (Beta, dif. units); unstandardized regression coefficient (coefficient B, units associated w/ measures)
cont.
(1) ethnography: study of ppl and cultures in a systematic manner; participant observation involved; dif. settings (2) issues: hard to get access to ppl (also ethnographic tours) (3) preserving info: triangulation: multiple sources to see if consistent narrative (better validity of info.) (4) grounded theory: bottom-up: theory from research; phenomenology: consciousness and direct experience (themes used to understand person's exp. NOT to create theory) (5) focus groups: (a) small gp, topic, leader focuses gp (b) similar participants, 5-10, preselected ppl, topic, wide range of perspectives, want discussion + have moderator, moderator asks questions, good gp environ (c) more efficient use of time; emancipatory pedagogy: learning approach where students and teachers are free from hierarchical roles (d) structure: intro, questions, discussion, closing questions
ch.8 qualitative research
(1) in-depth investigation: focus gp, interview, case study; no RA or manipulation (non experimental research); archival research, physical traces, participant and non participant observation, interview, case study, focus gp (2) reflexivity = researchers reflect on own exp. as part of research process; positivism: info. gained through senses; mixed methods = quantitative/qualitative combo; hypothetico-deductive method: scientific method: hypothesis + falsify (3) archival research/document analysis: no live participants; need de-identified data (HIPAA); physical traces: physically observe/analyze, absence of ppl, what's on chalkboard, trash on ground (4) observational methods: collect info via senses, naturalistic observation (observe in natural setting); internal validity problems (perception of what observing + what have observed in past) (5) behavioral categories: guides recordings; behavior coding scheme: review lit/what others did, pilot observations to see if categories work, adjust: add/remove categories (6) recordings: videotaping (leave camera), concealed observer (passive deception), acclimation/observation: in setting before study and ppl adapt; behavior recording/mapping: check box by behavior on sheet (behavior map = spatial record of behaviors/locations of ppl) (7) how often/how long observe: frequency (number of times), duration (how long lasts), interval (number of times in time period) (8) inter-rater reliability: percent agreement (number agreements divided by number events observed), intraclass correlation coefficient (association of ordinal/ratio V), cohen's kappa (nominal data, corrects for guessing) (all look at agreement + acceptable values) (9) covert observation: not informed purpose of being observed
from review session
(1) interpret t-test: The scores in gp A (M = 32.56) are significantly greater than the scores in gp B (M = 7.65) (t(22) = 7.99, p < .05) (2) interpret 1-way ANOVA (only 2 gps, 1 df): can say what direction (greater/less than); if more than 2 gps just say there's a difference (3+ gps; b/c need post-hoc) (3) ANOVA 2-tailed; t-test 1 or 2-tailed (4) population: kids 4-6 yrs and 11 months old with autism; sample: 22 kids 4-6 yrs and 11 months old with autism (5) quasi: chi-square or multiple regression (gped based on age and gender and looked at color-blindness; had gps not just looking at DV like correlational) (6) independent samples t-test: 1 IV and 2 levels; correlated measures/repeated measures t-test if measure multiple times
cont.
(1) interpreting interactions: statistically significant gene x gender x drug interaction (F(1,16) = 4.564, p = .0465); graphing 3 IV need 2 graphs (one male, one female; each with drug and gene); can only analyze 3 IV; always want to analyze gender (2) factorial design ex: drug x gender x gene interaction: 2 x 2 x 2; interpret interaction first (3) review of error types: (a) type 1 error: reject the null hypothesis when there is in fact no significant effect (false positive); p-value optimistically small (b) type 2 error: not reject the null hypothesis when there is a sig effect (false negative); p-value is pessimistically large (c) prozac vs. depression away w/ T2 error: approve it when not effective, kill ppl (d) with T1 error: lose out on something good (but needs to be as good as or better than what's on the market)
item/stem info
(1) item construction: (a) age-date of birth and current date (b) problems w/ using predetermined ranges: they're arbitrary (c) demographics later b.c sensitive and don't want drop-outs (d) want open-ended more, but harder to score (e) do open ended when possible/makes sense (f) gender/special issues do later on, can do prefer not to answer/not applicable (2) stem construction: (a) how satisfied/dissatisfied = leading question/build in bias (better/should do: how do you view...) (b) use scale of extremely dissatisfied to satisfied w/ neutral question (c) number items, bold numbers and stem
Methods/finding scales
(1) look at measures/materials/instruments section: (a) see what other authors used + if widely used/recent (b) psychometric properties: quantifiable aspects of measure indicating statistical qualities (c) Cronbach's alpha to see reliability (d) don't have all items; look in citations and go back until get to original (2) databases: (a) PsycTESTS: APA of tests w/ some items (b) HaPI: instruments w/ sources used in (c) PsycTESTS: see authors, number of items, response format, citation, permissions (d) detailed record: part of info that might give psychometric info (e) reverse scoring: state in manner opposite of other items + reversed anchors (stated neg when all other items are positive) (reverse numbers for group that is lower; ex. more high-self-esteem, reverse score low-self-esteem) (f) PsycTESTS = convenient/useful/see items; drawback: if look before lit review, not most widely used/reliable 1st
ch.9 between-subjects designs
(1) many conditions in a study, subject only receives 1 = between-subjects; mixed designs: wn and bet subjects components; no carryover effects (2) nomenclature by IV: (a) variable = independent variable (b) factor = IV/dimension from factor analysis (c) treatment = IV (d) condition: exp/control (e) level: version of trmt receive (f) value = variable/trmt level (g) trmt level = condition in exp; defined by levels: exp/control (h) all refer to IV (3) IV/factor/trmt/variable refer to IV; level/condition/value refer to which version of trmt/IV participant got (4) demonstration: illustrate variable impact, no control condition; independent samples t-test: two levels of 1 IV and one DV (exp and control gps); want adequate power/ES, partially accounted for by sample size (5) more than 1 IV: interaction effect: effect of one IV isn't the same for all levels of 2nd variable; simple effects: evaluate interaction components: impact of 1 IV is limited to one of the levels of 2nd IV (6) types: (a) randomized gps design: randomly assign gps to dif. levels of 1 or more IV (b) One IV/2 levels (independent samples t-test): levels = control/exp; look at means (c) One Way ANOVA: see if means of 3+ gps differ (1 IV with at least 3 levels, 1 DV) (d) Two-Factor designs: two or more IV; more than 1 IV = factorial design; fully crossed design (all combos of levels of IV are represented) (e) Two-way ANOVA (two independent variables)
correlational cont: simple and multiple regression
(1) marijuana use ex: (a) factors impacting use (b) recreational marijuana use can lead to cannabis use disorder (addicted to cannabis) (c) what factors make ppl use recreationally and might also lead to disorder (prediction model) (d) y1 = a + bX is simple regression (intersection pt on y-axis = a, b = slope, X = specific score) (e) simple regression: looking at 1-2 factors (2) multiple regression: (a) multiple factors; ex. tree planting, deer population, predator population (b) recreational marijuana use = gender, age of first use, location, reason (c) identify multiple factors/points to predict future outcomes (d) predict success in college = GPA, ACT, mental illness, motivation, family issues (e) DARE programs: increased drug use, but more likely to seek help/intervention (f) only works if correlation coefficient is large (3) qualtrics: (a) online platform, get large sample but soliciting not super effective (email, mail) (b) can recruit in classroom + have ppl take survey in classroom/lab (control environ take it in) (c) need to be anonymous to answer sensitive questions (d) timing matters too (CBD legal or not; being tested for drugs or not) (e) need follow-up with formal drug test + can't convict on first + many ppl sit in same spot
cont.
(1) matched group design: before RA, match on 1+ characteristics impacting DV; multiple comparisons: make several comparisons simultaneously to evaluate hypothesis (but T1 error likely) (2) error rate per comparison (PC): in any comparison, prob of making T1 error; family wise error rate: prob of having at least 1 T1 error in set of comparisons (3) planned comparisons: dif expect to see; unplanned: no advance predicts/run comparisons post hoc (but higher T1) (4) error variance: (a) variability in score not from IV, not systematic, not controlled (b) how address: RA to conditions, increase sensitivity of IV and DV, be consistent (treat the same/behave/appear the same to all participants) (5) considerations: sensitivity/level of IV, number of IV, number of participants, power/sample size, interaction effect interpretation, multiple comparisons/T1, error variance (6) multivariate analysis of variance (MANOVA): (a) 2+ DV conceptually related (b) combo effect on IV (c) limits: multicollinearity (DV highly correlated/redundant, not unique contribution) + outlier (far away data point from error/variability in msmt/chance)
ch.5 cont. in class
(1) maturation: (a) if 120 items or 35 pg with 30 items/pg, no one going to answer + if do, super dissatisfied (b) not reliable or valid (only ppl who answer are those angry at company) (c) participation depends on fatigue or hunger (d) need measure long enough for goals + short enough to avoid maturation effects (2) question construction: (a) strongly agree, agree, neutral, disagree, strongly disagree (b) 0-4 or 1-5 (c) orange x. in class: hard to give 0 and strongly agree (for 0 hesitate and give higher rating so don't have to give a 0) (c) what expect impacts questions, ex. expect neg. results (d) scale w/o middle/neutral: bad, lots ppl neutral and will go like/dislike when actually neutral; should use a neutral
Amaral paper 2/2
(1) more white matter than grey matter: (a) theory autism vs. typical: more white matter than grey matter (b) gestation: least white matter + lots grey matter at 6 months; can't have myelin of axons w/o connections (brain not functional until connect neurons) (c) myelin = white matter; more connections; lose grey matter and replace with white/connections = typical (d) autism more connections early on; need to give education/teaching at right intervals (2) more: (a) or could be glial scarring: more white matter (b) more white matter from myelination or glial scarring (c) prob not glial scarring, but alt. hypothesis to consider, want to set up to do null hypothesis testing (d) 14% reduction in white matter in corpus callosum (connecting hemisphere) (3) corpus callosum: (a) 3% typical/gen pop don't have connected hemispheres; 7% autistic individuals (A colossal)
ch.7: nomenclature
(1) nomenclature: (a) ranked data: ordinal, something higher/more preferred; not equal distance bet them (b) order of preferences (2) nominal/one dimensional chi-square: (a) goodness of fit: population as whole; dif. bet sample and hypothesized distribution (3) preexisting gps: ANOVA dif bet means; same info for quasi-E as E, but no causality claimed (4) two-dimensional chi-square: two nominal dimensions (4) gp differences in ranked data: sample as a whole or compare ranks for dif. gps; ranks based on ordinal data (Pearson's r = interval data); Spearman's rank order (rho) (correlation for nonparametric) (5) factor analysis: take items, cluster gps of items to differentiate maximally; factors = clusters of items that are different from others (use factors as DV, ex. types of buildings: traditional house, brick office, large medical)
Operational definition/construct + validity
(1) operational definitions: (a) specific in methods section: define what measuring; doing in experiment + what construct is (b) construct: emotional/mental states can't see it; but can measure (c) ex. anxiety = mental/emotional state, symptoms in DSM-5, anxiety construct + how operationally define (d) learning = test performance (e) construct: define this, then create objective/valid/reliable measures (2) validity: (a) accuracy, does measure what it's supposed to measure + based on standard (b) does midterm assess what really know; is A valid indicator of learning, ability to get job, go to grad school (c) A - succeed in grad school; C - don't succeed; if both succeed prob not good measure
scales cont.
(1) ordinal scale: (a) order, don't know distance of one choice in order from another (b) rate transportation 1 to 7; see order but don't know if level of preference varies (c) Spearman's rho: correlation bet rank sets + Kendall's tau: inversions in ranks bet ppl (2) interval scale: (a) widely used, continuous data (b) anchors equally spaced; 5 pt scale: same distance bet 3 and 4 and 1 and 2 (c) use parametric/ANOVA if meet assumptions (normal pop distribution); nonparametric if don't (d) Likert response scale: 5 anchors (points on rating scale) = strongly agree, agree, neutral, disagree, strongly disagree (e) can have 0, but it is relative NOT absolute value, like 0 on scale (means something) (3) ratio scale: (a) a true 0 (interval doesn't) (b) no correct answers, no weight/height (c) 0 = absence of quantity (4) scale-sensitivity: (a) ability of scale to detect differences (b) 7 pt scale better than 3 pt scale, can have more varied answers (5) measures = DV; stimuli = IV
trace evidence ex in class
(1) overall themes from sample: electrical, repair/construct with bolts/screws, microscope slides, office space: graph paper/paper, tissues and cleaning wipes (2) lab feel but not university; maybe robotics, clinic; need to analyze what's on tissues and microscope slides to know for sure (3) acceptance of qualitative evidence/reaction: (a) APA founded in 1892 (b) psychonomic society founded in 1960 - APA diverts from experimental methods (c) 1970 society for neuroscience founded (d) 1982 APA creates divisions in an attempt to stop the loss of experimentalists (52 or 53 divisions) (e) 1988 APS formed (f) 2000 qualitative research (added to APA) (g) 2008 Journal of Mixed Methods (qualitative research journal)
ch.10: within subjects/mixed/pre and post designs
(1) participants receive all conditions; repeated measures for same ppl; more power w/ fewer ppl; act as own controls (2) prepost designs: msmt before and after same intervention (3) range effect: effects of multiple exposures -> context effects: practice effects/sensitization (susceptible to stimulus)/carry over (differential carryover effect = lingering effects are different from 1 V to another) (4) maybe more external validity; subject mortality: loss of power when ppl drop out b/c in all conditions (5) counterbalancing: (a) control for carryover (b) complete counterbalancing: all possible orders are used; partial: some but not all possible to control for order effects: random order dif. for all participants; Latin Square: each trmt in each row and each column one time (c) ABBA order: presented in one order than its reverse (AB, BA) (d) randomized block design: participants exposed to multiple conditions more than once (6) add complexity: more levels or another variable (7) can also do wn-subjects MANOVA (2 related DV) (8) mixed designs: (a) wn and bet components (b) randomized bet intervention + wn assessment (9) pre-post: (a) measure, intervention, measure (b) RA to condition (c) single gp: no control gp, measure b4/after (d) exp-control: add control gp (e) Solomon 4-gp: some pre/intervention/post, pre/no intervention/post, intervention/post, and intervention/pre
Surveys: developing measures + items
(1) pick existing measure, know reliable and valid; why don't pick existing: (a) none assess construct/want to improve existing (b) cost (c) need qualifications (d) doesn't address area of interest (2) Schwartz: (a) see -5 to +5 scale different than 0 to 10 (bipolar vs. unipolar) (b) do what's socially desirable, interpret middle as typical; ends = extreme (c) use open-ended: how many times a day do you text? ______ times/day (need unit) (3) demographics: (a) age/gender/race/edu/income/ relationship (b) sensitivity: let self-define: Gender: ______ (fill in the blank) (c) open-ended = specific (d) race/ethnicity: open-ended, race (bio) ethnicity (culture/geo); for representative sample/generalizability (use app. specificity + respect ppl's language to describe self)
Cont.
(1) pilot test questions, increase internal validity (fix errors) (2) interval = parametric stats, widely used, more sensitive DV (3) nominal: (a) yes/no (b) content analysis: open-ended res, create categories/capture themes (c) mention category or not (course or not) (d) interval scales: degree of ___ (attitude/perception) (4) Dillman + question order: (a) questions abt research first (b) demographics last (c) later on: build up to sensitive items (5) online survey tools: (a) SurveyMonkey, Qualtrics (b) anyone/anywhere can do it (c) AmazonMTurk: crowdsourcing (obtain ideas/services from large gp of ppl) (paid to take survey) (d) more accuracy/less error, sustainable, saves time (e) disadvantages: cost, learning curve, some don't have internet access, flexibility -> threat to internal validity (not taking survey in same conditions) (f) can do survey in-person
power
(1) probability of statistical test correctly rejecting the null when the alternative hypothesis is true; should be done before the start of the study (2) 1-B, where B is probability of making a T2 error (3) ability to find an effect if the effect exists: daphnia's behavior; lecanemab treats Alzheimer's (4) power related to sample size and effect size (5) power analysis: stat power one piece of puzzle with four related parts: (a) effect size: quantified magnitude of result present in population; calculated with stat like Pearson's correlation coefficient for relationship between variables or Cohen's d for dif bet gps: chi-squared, phi, r^2 coefficient of determination (b) sample size: number observations in sample (c) significance: sig level used in stat test; ex. alpha 5%/.05 (used in most except toxicology) (d) statistical power: probability of accepting alternative hypothesis if it's true (6) ES and power largely estimated; only know if replicating a study; for unknown ES: want largest sample size possible; ES related to SS/sig/power (7) can estimate ES if doing existing compound vs. new (aspirin w/ 150 yrs of data)
Amaral paper 1/2
(1) problems: (a) can't diagnose autism until 2-3 yrs (b) but neurodevelopment disorder + develops over time; greater brain volume/greater head size so accelerated brain development (c) critical period theory: has to occur within a certain window (d) pruning: brain removes unnecessary senses; if don't acquire skill/behavior before pruning, can't ever acquire it (ex. language) (e) autism: brain grows so quickly miss windows; savant: good in a certain area; earlier window than others b/c of accelerated growth (in 4-7 yrs vs. 25 yrs) (f) compressed into small window -> teach in window -> normal development (16 normal size, growth stops) (g) early yrs progresses quick, 1 view = accelerated growth and miss window if diagnose at 2/3
questions + validity
(1) questions: (a) open-ended: put anything down, write as much as want to (b) closed-ended: get options, easier to score/analyze b/c fixed options (c) open harder to score but get more info. (d) instead of asking question about if graduated school/or if can see colors; do a task: math problems or actually differentiate bet two color samples (e) give scenarios: what do in this situation (f) employment test 3 types of questions: mathematical skills/personality/product knowledge (2) validity: (a) face validity: if can detect color, can recommend color (are you able to see a wide variety of colors?) (b) predictive validity: look at returns, what sold, mistakes at register (test predicts employment/do job performance eval 1 year later) (c) concurrent validity: give test to someone else, like current employees + compare test scores to how currently performing in job
qualities of measures: reliability
(1) reliability: (a) repeatability and internal consistency (b) consistency measured over time (test-retest) OR refer to degree items measure same construct (internal consistency) (c) test-retest: instrument given at two points in time + scores = correlated; use Pearson's R (d) ex. students who took same self-esteem measure at two points in time (2) parallel-forms reliability/alternate forms reliability: (a) take SAT twice to improve score (2nd time, parallel form) (b) inflate scores b/c familiar with items, need other items (c) equivalent forms of test are shown to be reliable (solves the issue) (3) measures of internal consistency: (a) split-half reliability: instrument split into halves and halves correlated w e/o (b) ex. test randomly divided into halves, take both w/ intermingled items at same time, then separate halves and see correlation (c) Cronbach's alpha: most widely used; should be .8 or higher
ch.7: specialized non experimental designs + longitudinal/cross-sectional
(1) time series (observations of same V over time); interrupted time series (msmt before and after something occurs); interrupted time-series analysis (multiple assessments before and after event; ex. legislation) (2) strength: ecological validity; weakness: threat to internal validity (no control gp, how compare) (3) postoccupancy evaluation (POE): building performance eval, posttest only; not experimental: no RA/manipulation (4) nonequivalent control group: no RA, pre-post sign, control gp like intervention gp used (5) longitudinal and cross-sectional: good when looking at development (6) longitudinal: (a) more sure changes associated with maturation (disadvantage: time/money, experimental mortality, repeated measurement might impact responses) (can do longitudinal + experimental; uses cohorts)
quality of measures: validity
(1) validity: measure assesses what claims to assess (measure of stress assesses what ppl think stress is) (2) content validity: (a) focuses on representativeness on domain of interest (spelling test for fifth graders composed of words selected from books read by fifth graders) (b) test has questions on information went over in class (c) adequately sampled from the domain of interest (3) face validity: (a) measures subjectivity appear to assess what you claim (measure of leadership asks abt decisiveness) (b) looks like what claims to assess (c) identify classical music, should play classical music (d) if mismatch, might increase demand characteristics
cont.
(2) more: (a) need longitudinal study; criticism: using cross-sectional when should be using longitudinal (but expensive, and often killed before conclude) (b) current longitudinal studies have very small sample sizes (c) diagnosis at 2-3 yrs (d) very small samples (e) cross sectional instead of longitudinal (f) older and high functioning in samples (skewed: might have dif. brain pattern than everyone else on the spectrum) (but practical, easier to get to stay still in MRIs) (3) comorbidity with multiple disorders: (a) does autism cause disorders (b) multiple related factors produce both (c) it's by probability/chance have these disorders (d) prob. combo of all (4) more: (a) need pure model to study autism; can't study both at once b/c interaction (change from which disease) (b) use animals to learn abt humans + can study from birth
cont.
(3) HaPI: (a) unique instruments number diffterm-3er from total (b) sort by relevance/author/source/date (c) see articles scale has appeared in + find original source; PsycTESTS: detailed info (4) books of measures: ex. Measures for Clinical Practice and Research: A Sourcebook; ask department/professors (5) catalogues w/ fees: (a) Psychological Assessment Resources (PAR), Consulting Psychologist's Press, Myers-Briggs Type Indicator (most well-known), EdITS, Western Psychological Services, The Psychological Corporation, Mind Garden (students use this) (b) only avail through company + need training, check PsycTESTS first (c) don't search online and use measure: unethical + might not be correct
experimental research 2/2
(3) One IV-One Way ANOVA: (a) if 2 gps, independent samples t-test (1 or 2 tailed)/1-way anova (2 tailed) (b) t-test b/c 1-tail (c) 1-way ANOVA w/ > 2 gps priori vs. post hoc comparisons (d) assumptions: normal distribution and homogeneity of variance (bet gps similar variance) (e) sample size and ES for determining group's N (4) between gps: two-way/two-factor ANOVA: (a) lab 2 example: emergence latency based on gender and gene (b) main effects first: gene alone; gender alone; then interaction effect: gene x gender (c) 2 x 2 ANOVA (2 variables each with two levels) (d) gender: male/female; gene: ASD/C57
cont.
(3) criterion validity: (a) standard: succeed in grad school/job/complete study at WSU (b) predictive validity: can test score predict if successful in job/grad school; does it predict future success/failure (b) A predict success/C failure/less success (c) concurrent validity: some standard ppl accept as legitimate/perfect evaluation of performance (GRE - get into grad school + see if grades measure up with that standard) (d) more ex concurrent: A on GRE and good grades = concurrent (what teaching in class matches with GRE); score consistent with something accepted as a c=perfect score; GRE = gold standard/current standard (e) predictive = future; concurrent: standard have to see if measure works (f) used to give GRE to WSU students stopped b/c regression to the mean (97/99%, can't do better) concurrent: did well on GRE and courses
cont.
(3) item order: (a) counterbalance res order within survey halves (10 1 format, 10 in another format); change construction so counterbalance choice order (ex. flip scale) (b) check for leading questions (c) construct validity (d) location of demographic questions (towards the end) (e) location of sensitive questions (towards the end) (f) recommended scale size: 5 or 7 (4) format: online, computer, paper, oral interview phone or in-person (5) number items, bold stem, standard font anchors, bold choices within anchors when in a line, otherwise use standard font
cont.
(3) picking stats: (a) what is the distribution? normal = parametric; skewed = nonparametric (b) might use one end of distribution, makes sense it's skewed (c) how collect data and how set it up/hypothesis testing impacts stats, too (d) want to test means/inferential, so do ratio or interval (4) overall: (a) what is distribution? normal or skewed (b) scale of msmt using (c) hypotheses? casual/C+E or correlational (5) more: (a) sometimes question set up so only can do correlational; if experimental: be careful with ethics (b) nominal = mode, no inferential states + correlational (c) survey creation: marijuana legalized: what's important?
cont.
(3) within subjects and repeated measures designs: (a) wn subjects = each subject receives each of the trmt conditions (no bet gps factor) (b) repeated measures - each subject receives the same trmt multiple numbers of trials (c) wn = all ppl receive placebo and 3 drug trmts, act as own controls; 20 subjects each trmt for human subjects (d) get rid of error variance and less subjects needed in wn (e) problems: carryover effects: effects of 1 trmt carryover to another (ex. long-term acting drug for trait anxiety for 6 months; each trial needs to be 6 months + 1 day apart to control for carryover); practice effects + maturation included (4) solutions to error types: (a) complete counterbalancing: two trmts 2 x 1 = 2 dif trmt orders; three trmts 3 x 2 x 1 = 6 dif trmt orders; five trmts 5 x 4 x 3 x 2 x 1 = 120 dif trmt orders (b) partial counterbalancing (c) ABBA: in case studies for problems w/ multiple factors ABAB/ABBA/BAAB with two trmts: AI not AI counterbalance across conditions + randomize in each subject for 2 dif factors (d) Latin Square: repeated measures; compare varieties of crops on same field broken into quadrants: water one factor and soil another; counterbalance 1 2 3 4 order throughout
cont.
(4) criterion-related validity: (a) test predicts outcome/criterion of interest (b) correlation of test score and outcome/criterion score (c) predictive validity: measure predicts outcome in future (test at one point, outcome at another point in time) (take SAT and ACT, then measure GPA later once in college) (d) concurrent validity: correlation of test and outcome occurs at same time (take test and performance/outcome measured at same time: give screening test to current employees and see correlation with their success) (e) restriction of range: distribution of scores not widely dispersed across range interest, impacts correlation size
cont.
(4) income: most ppl don't like this one: don't know or invasive (5) item format: (a) anchors and stem -> stem = statement/question/prompt ppl reply to (b) 5-7 anchors; if odd, can be in the middle/neutral (c) don't want leading stem: how satisfied do you feel (lead) vs. how do you feel (d) stem bold, res not
cont.
(4) video: (a) dark room 0 lux; rodents = light phobic (nocturnal, afraid of light); sit in dark for 30 sec, then in light chamber and then back into dark (b) want to explore light-dark test of anxiety: anxiolytics (reduces anxiety, Valium, barbs) and anxiogenics (induces anxiety) (c) autism should produce anxiogenic -> more anxious (afraid of light naturally, autism increase anxiety, more anxiety behaviors; probs won't come out of dark room at all/take longer) (d) emergence latency: amount of time to enter (standard DV); half placebo and half valium; come out in 10 sec (valium) and 300 sec (control) (e) want moderate anxiety (can't detect effect if on extremes, ex. floor/ceiling effect) (f) control = middle anxiety strain (fear of light, but curious; worse anxiety -> stay in dark)
cont.
(5) discriminant validity: (a) situation in which measures hypothesizes to be theoretically unrelated are in fact unrelated (6) construct validity: (a) measure assesses what it claims to measure (b) nomological network: lawful relations need to discover to validate a construct (c) relationships of variables theory predicts: shoe construct correlates with similar measures (convergent) or does not correlate with dissimilar (discriminant)
cont.
(5) nominal scale: (a) no ordering, produce categorical data (put into categories) (b) yes or no = ex., doesn't matter order just count (c) open-ended res usually (answer any way) over closed-ended (select res from provided options) (d) measured w/ frequencies (count number of res in each category) or chi-square: nonparametric, see if distributions of dimensions differ (1st vs 4th yr HS students on sports/academics/social life) (6) parametric: make assumptions about pop fitting normal distribution; nonparametric: don't/no assumptions (nonparametric less powerful and need bigger samples sizes to increase power)
cont.
(5) operational definition: (a) emergence latency: 4 paws or 2 paws out; clear operational definition for good inter-scorer reliability (so use same definition) (b) put operational definition in procedure section (c) constantly shifting bet dark and light: compare time in dark vs. light - inverse (same measure: time in light = DV, valium spend more time in light/autism less time; time in dark: non-autism more; autism less (d) others: where in box (on edges or center). activity level (want 2 DV)
cont.
(6) SurveyMonkey/Qualtrics: (a) have to pay or do upgrade, sometimes don't export to SSPS, lots limitations (b) features: change anchors + number, progress bar, randomizer function = distribute conditions/questions randomly (7) Google Docs Forms: survey data, res saves to spreadsheet + free; download into SPSS: have to add variables (or just label them in survey initially) (8) survey appearance: (a) bold stem, not anchors; use spacing (b) add question numbers; can add page breaks + prompt if missed question (c) care more: ppl take it seriously + better internal validity (ppl res to more questions)
cont.
(6) interviews: (a) done w/ an individual; more info, but take time/resources (b) structured interview: items prepared in advance, same questions asked in same way to all ppl (c) unstructured: plan, but no certain questions; use open-ended (d) semistructured: set questions guide, but room for intro, follow-up, related questions (d) interview schedule = set questions (e) train interviewers but also interviewer effects (who of the interviewer, their demographics) (7) case studies/histories: (a) records of info. of ppl's medical/psych history = case histories (b) ideographic = focus on individual + individual exp. (c) nomothetic: develop laws of behavior to apply generally (d) need nature of case (activity/functioning, background, setting, context, other cases, informants) (8) content analysis: (a) analysis of res; qualitative and quantitative meet; want open-ended res from open-ended questions; read through/make categories/operational definition of category (b) steps: read through, make res list, categories list, operational definition, inter-rater reliability (9) content categories/stats: you choose or CAQDAS; research = emotional
cont.
(6) method: participants/who, materials/with what, procedure/how (7) maturation: fatigue from number of scales and number of items, select carefully (threat to internal validity) + Flesch-Kincaid scale: provides stats for ease of reading + need to score (may be sub scales, too) (8) names and social desirability: (a) name should communicate content or measures with unwanted expectancies/demand characteristic (part of research process unintentionally influences ppl's res) (b) if know what assessing, change answer; do what's socially desirable (perceived favorably) (c) covariate: V that can effect relationship of DV and IV being addressed (reduce error) (d) Implicit Associations Test (IAT) use rxn time and see associations bet concepts, reduce social desirability (more associated, quicker res; ex. old good vs. young good)
cont.
(b) corpus callosum shares info bet hemispheres; materialized brain with logic and emotion but connected and need both to function (c) historically, SZ/psychotic had lobotomies/separated hemispheres (disconnected emotions) (became Zombie-like; changed behavior) (4) more: (a) disconnected from birth might show neurons not growing where supposed to (b) neurons in hippocampus, depends on where look: reduced/missing in some areas and a lot/packed in other areas (c) class-shape is altered; dif. distribution of neurons, neurons are in the wrong place (d) packed in dif spots, same # neurons in dif. places (5) more: (a) 3 core features of autism: social impairment, communication deficits, repetitive behaviors (b) structures changed, migration/# of cells = dif. (c) 23 brain structures w/ dif. migration; compact in 1 area + reduced/missing in others (d) something wrong/dif. with cues