RM1 Step 6: Select a Research Design (pt 1)

Between-subject Design

Between-subject design (BSD)

Research by comparing scores across different individuals from separate groups 

Advantage of BSD

Each group is only measured once, so participants’ scores are unaffected by order effects:

(more on the next chapter)

Disadvantages of BSD

1. Requires a large number of participants

2. Vulnerable to individual differences within groups: people differing in personal charactersitics

What 2 problems can within-group differences cause?

1. Potential confounding variable —> threaten validity

2. Produce high variability/variances in scores, making it hard to see the treatment effect

Potential confounding variable

Recap: this can be solved using random assignment

High within-group variance in scores

Within-group variance = error in inferential statistics = ‘noise’, making it hard to see a clear difference BETWEEN groups

*Note: random assignment will not solve individual differences WITHIN groups, only between groups

How does within group variance threaten validity?

High within group variance can mask the effect of the treatment condition eg. chocolate-insomnia

Goal of BSD

Maximise between-group variance, and minimise within-group variance; people across groups to perform differently, and people in the group to perform similarly 

4 Ways to minimise within-group variances in BSD: (LSIU)

1. Limit individual differences

2. Standardise procedures and treatment setting

3. Increase sample size

4. Use Within-subject Design (next chapter)

Limit individual differences

Holding a variable constant or restricting its range (eg. gender, age); only sampling participants with a specific characteristic

(-) 

BUT not popular as it could reduce EV

Standardise procedures and treatment setting

Ensure all participants are treated the same

Eg. use the same experimenter as individual differences could arise from differential treatment

Increase sample size

A large sample can statistically overcome high variance

(-) 

BUT very costly: the influence for sample size occurs in relation to the square root of the sample size (eg. to reduce the effects of high variance by a factor of 4, you need to increase the sample size by a factor or 16)

Use within-subject design

Last resort if within-group variances are extremely high