Correlational Approach
Eron et al. (1972) (Longitudinal design)
- two separate measurement time periods for 211 male participants (3rd grade and 13th grade) - predictor variable: amount of violent TV viewed - outcome variable: peer-rated aggression levels - TV violence seems to come first
Cross-Sectional Correlations
Correlations between two variables within a time point (TV violence 3rd grade) <--> (Aggression 3rd grade) related and significant
Internal Validity
- Can we make a causal claim in correlational studies? NO - Internal Validity is low
Rules for Determining Causation
- Covariance - Temporal precedence (one variable coming before another) - Internal validity - correlation does not meet ALL causation rules --> therefore correlation does not imply causation
Sanbonmatsu et al. (2013) (Correlational Approach)
- Is multitasking ability related to how much time people multitask AND is multitasking ability related to perceived ability? - predictor measure: multitasking ability (using OSPAN) - outcome measure 1: how much time a person actually multitasks (using MMI (media multitasking ability)) - outcome measure 2: perceived multitasking ability - people who score high on OSPAN should be good at multitasking - people are bad at judging their multitasking ability - relationship between OSPAN and MMI: r = -.19, p < .05 (small, inverse relationship; statistically significant) - relationship between OSPAN and perceived multitasking ability: r = .08, p > .05 (NOT statistically significant)
Mehl et al. (2010) (Correlational Approach)
- Is small-talk related to happiness? Is the amount of substantive conversation related to well-being? - 79 undergraduates wore the EAR (electronically activated recorder) for 4 days; recorded 30 seconds of sound every 12.5 minutes - predictor measures: 1) alone or talking to others 2) small-talk or substantive conversation (relational) - outcome measures: 1) well-being 2) satisfaction with life 3) happiness - outcome: small-talk IS related to well-being - people who have more small talk are LESS likely to be happy -people with more substantive conversation are MORE likely to be happy
Cacioppo et al. (2013) (Correlational Approach)
- Is where a couple met (online or offline) predictive of whether the couple stay together in the future? - 19,131 individuals married between 2005-2012 -predictor measure: met online or offline - outcome measure: whether or not participants are still married, separated, or divorced - significant relationship - more likely to stay together if you met offline (SMALL relationship, but statistically significant) -number of participants influences results as significant
External Validity
- To whom can the association be generalized? -mostly use convenience samples in correlational studies - not super concerned with external validity
Correlational Approach
- all about measurement/measured variables - looks at association between variables - cannot make causal claims -CONSTRUCT validity and STATISTICAL validity are important to the correlational approach -can't have internal validity
Woehr & Cavell (1993) (Multiple Regression)
- each predictor variable has 3 components -what factors influence academic performance? - predictor variables set 1: academic ability 1) high school GPA 2) SAT verbal score 3) SAT quantitative score - predictor variables set 2: academic effort 1) time studying 2) chapter readings 3) classes missed - predictor variables set 3: nonacademic activity 1) school-based extracurricular activities 2) work 3) television viewing - outcome variable: General Psychology Test score
What are some strategies that researchers can use to address limitations with the two variable correlational approach?
- longitudinal design - multiple regression
Construct Validity
- measurement - reliability -validity
Factors that influence correlation coefficient in an artificial way - Outliers (Statistical validity)
- outliers = extreme scores - outliers can artificially increase correlation coefficient (bigger than it should be) - outliers can also artificially decrease correlation coefficient - can identify and remove outliers to get a more accurate correlation coefficient - outliers have a greater effect on the correlation coefficient when you have a smaller sample
Curvilinear relationship (Statistical validity)
- the correlation coefficient is designed to detect straight lines - NOT curves
Statistical Validity
- what factors influence our conclusions about the data? 1) size of correlation coefficient 2) statistical significance 3) factors that influence correlation coefficient in an artificial way (restricted range) 4) is the relationship linear?
Factors that influence correlation coefficient in an artificial way - Restricted range (Statistical validity)
- when you don't collect the entire range of variables --> underestimate the correlation coefficient - can decrease correlation coefficient - we don't want restricted range
Size of the correlation coefficient (Statistical validity)
0.10 - small, or weak effect size 0.30 - medium, or moderate effect size 0.50 - large, or strong effect size - the bigger the effect, the better I can predict one score from another - generally like/prefer/want studies with larger effect sizes - small effect sizes can be important under certain circumstances - e.g.: medical
Statistical significance (Statistical validity)
1) effect size 2) n = number of people in a sample (a large sample and influence the significance of results)
Cross-Lag Correlations
Assess whether the earlier measure of one variable is related to the later measure of the other variable - helps with temporal precedence (TV Violence 3rd grade) <--> (Aggression 13th grade) significant moderate effect
Individual Predictors - Betas
Betas = index of relationship between one predictor and the outcome while controlling for the other predictors (correlational coefficient in regression model when other variables are controlled for)
Autocorrelations
Correlation within a particular variable between time periods (Aggression 3rd grade) <--> (Aggression 13th grade) significant, moderate effect
OSPAN Measure (measures multitasking ability)
Instructions for OSPAN measure 1) read math equation out loud 2) determine if equation is correct or incorrect (say out loud) 3) read the target word out loud and try to remember it for a later recall test - people on average can correctly recall a 4 word sequence
Woehr & Cavell (1993) (Multiple Regression - Intercorrelations and multiple R)
Intercorrelations: - Time studying = 0.01 (no correlation, not significant) - Classes missed = -.0.22 (statistically significant) Multiple R: -R = 0.50 - R is significant
Temporal Precedence
One variable coming before another
Combined relationship or multiple R
Similar to correlation coefficient but does not take on negative values - how well do all predictors predict outcome?
Multiple Regression
Statistical technique where you can assess multiple variables and look at their relationship to one outcome - internal validity: CAN rule out 3rd variable - each circle = one variable - overlap = shared variability/relationship - can take out one circle and just look at overlap between other circles/variables
Longitudinal Design
When researchers measure multiple variables at different times; measurement over time - cannot rule of 3rd variable influence