Week 4-T tests Flashcards
Give 3 different names of within-subjects t-tests
1.Paired samples
2.Dependant samples
3.Related samples
What are within-subjects t-tests?
Compares the mean of the same group but at different times e.g., before vs after treatment
Give 4 different names of between-subjects t-tests
1.Unpaired samples
2.Independent samples
3.Unrelated samples
4.Alternative:Welch’s test
What are between-subjects t-tests?
compares the mean of two independent groups (one group of participants compared to another)
What’s the history of the student’s (Independent) t-test?
■ William Gosset was a chemist who worked in the Guinness brewery in Dublin
■ Created the t-test as an easy way to measure the stout quality.
■ Guinness banned their workers from publishing findings.
■ So published under the name ‘student’
■ T-tests sometimes called ‘Student’s t-tests’ BUT now mostly referred to as the ‘Independent t-test’
What design is the Independent t-test?
Between-subjects design
What are the 3 assumptions of the independent t-test?
■ Data must be continuous / scale (i.e., Interval or Ratio)
■ Data must be normally distributed (i.e., not skewed)
■ Data must have homogeneity of variance (i.e., variance between groups are equal)
True or false: comparing 2 different groups can be naturally occurring groups: sex (male vs. female), age (20s vs. 50s) OR experimental groups: treatment type (drug vs. placebo).
True
Define heterogeneity of variance/heteroscedasticity
The variance between groups is not similar
How do we know if we have homogeneity of variance?
■Levene’s test: Tests the null hypothesis that the population variances are equal.
■Levene’s test is significant (p < .05)=HoV is violated. (and is hetero)
■Leads to an increased likelihood we will find a significant effect when there isn’t one (false positive; Type 1 error).
Do you need to report Levene’s test in a report?
NO. You just report the results of the t-test with either equal variances assumed or equal variances not assumed.
What’s the Welch’s Test?
■ An alternative to the Independent t-test which doesn’t assume HoV.
■ It’s actually an F-test which technically makes it closer to an ANOVA, but it can be used for two groups.
True or false: Some statisticians suggest we should always report Welch’s test regardless of homogeneity of variance
True
What design is a paired-sample t-test?
Within-subjects design
What are 2 assumptions of a paired-samples t-test?
■ Data must be continuous / scale (i.e., Interval or Ratio)
■ Data must be normally distributed (i.e., not skewed)
