Title | Unrelated t-test |
---|---|
Author | Jaxson Mannes |
Course | Design and Analysis of Psychological Iverstigations |
Institution | Goldsmiths University of London |
Pages | 2 |
File Size | 96.8 KB |
File Type | |
Total Downloads | 19 |
Total Views | 153 |
Notes on Unrelated t-tests....
Notes Unrelated t-test
Assumptions of data for Parametric Tests o Must be at least interval
Interval-On scale
Ratio- On scale with absolute zero
Also ok
o Sample drawn at random from population o Sample data should come from normally distributed population
Does not mean that sample data should be normally distributed
Most samples are too small to assess whether they are normal
In practice- assumed population has a normal distribution
Decision may be based on past experience or theory
Unrelated t-test o Looks at differences between 2 conditions or groups o Must have interval level data o Must be between groups design o Test-independent samples on SPSS o Data Assumptions
All of the above Parametric assumptions
Homogeneity of variance
Variances of groups should not be significantly different
Need to check when… o Using an unrelated design o Sizes of groups are different o SPSS does this automatically-Levene’s test
Robust test- some leeway in assumptions
Normal distribution- Some flexibility
Homogeneity of variance (tested by SPSS)- Alternative results to report
Data must be interval
Alternative- Run a Mann-Whitney
o Reporting the test
T(df)= t value, P=significance level
Format is the same for both related and unrelated t-tests
Ex. T(26)=2.79, p=.01
Ignore signs of values, always put in as positive...