Lecture 14 Mixed Designs PDF

Preview

Summary

Instructor: Karen Lawson...

Description

Mixed Designs ● = a type of factorial design ● = A combination of between-subjects and within-subjects designs ● Has more than 1 IV, has main effects and interactions ○ One IV = a between-subjects variable ○ Another IV = a within-subjects variable ● Example: longitudinal study on impact of early specific skills training on children’s IQ Age 3

Age 4

Age 5

Age 6

Age 7

Age 8

Social Skills Cognitive Skills Control Group ○ ○

●

DV = IQ score IV = skills training ■ Between-subjects variable - kids are assigned to each group at a time ○ IV = age / time ■ Within-subjects variable - all kids will be in all treatment conditions (age 3, 4, 5, so on) ○ Examine main effect of skills training, main effect of time, interaction effect Example: sex differences in children’s fear of the dark Dark Room Scary Image

Dark Room Neutral Image

Lit Room Scary Image

Lit Room Neutral Image

Boys Girls ○ ○ ○

DV = fear / heart rate IV = stimulus type ■ Within-subject variable: all kids will be exposed to all stimuli IV = gender ■ Between-subject variable: each kid can only be in one group: male or female

Matched-Subject Designs ● Matching the participants b4 they are assigned ● Characteristics 1) Each participant is exposed to only one level of the IV (inherently, a between-subject design) 2) Each participant has a matched participant in each of the other conditions so that the groups are correlated / not entirely independent 3) Only one DV measurement per participant, but the analysis takes the matching into account 4) Critical comparison is the difference between the correlated groups

●

●

●

●

●

When to use: ○ We want the sensitivity (suspect a small effect) and there’s potential carry-over effect (can't use within-subject design) ■ Between-groups design: random assignment to independent groups but not as sensitive ■ Within-subject design: sensitive to small treatment effect, but cannot use if there’s strong carry-over effects or excessive demand on participants Matching participants ○ Pre-identify one or two most salient confounds (from past literature) ■ More variables we need to match, the more complicated ○ Match participants in sets, where the size of the set = # of conditions/treatment groups ○ Once the sets are matched, randomly assign the participants in the set to the conditions Advantages ○ Increased sensitivity to small treatment effects ○ Requires fewer participants - increase power ○ Avoids order effects Disadvantages ○ Extra work of matching participants ○ Loss of potential participants without appropriate matches ○ More complicated situations: ■ More than 1 confound identified, more to match ■ Difficult to match on continuous variables (eg. age, we need to use age groups, 20-30, 30-40, etc) ■ Large number of treatment conditions, more difficult Example: Which is more effective for treating depression: psychoanalysis or CBT? ○ IV: psychoanalysis, CBT; DV: measure of depression ○ Expect to be a small treatment effect ○ Match groups on potential confounds: age, gender ○ Pair participants according to confound (1 male 30-40 paired w/ 1 male 30-40; 1 female 20-30 paired w/ another female 20-30, etc) ○ Randomly assign one of the pair to each group (1 male sent to CBT, the other sent to do psychoanalysis) ○ MR Group 1 CBT BDI MR Group 2 Psychoanalysis BDI ○ We now have 2 groups matched on age and gender (they are equivalent on lvls of confounds - reduced error), we can now compare 2 groups on the DV ○ Can get more info by analyzing the matched factor (age, gender): ■ Confounds = unintentional IVs, contributing to changes in DV but not intended by us ■ We can make them an intentional IV, which turns our study into a factorial design ■ Examine the main effects of IV & confound, as well as interaction effects

■

Gender

CBT

Psychoanalysis

Men

BDI

BDI

Women

BDI

BDI

Now gender (a previous confound) is another IV, we examine the main effect of therapy type, main effect of gender, and interaction effect of gender & therapy type...