T Tests Type I Error: incorrectly finding a significant effect – you could only avoid this through testing the entire populations because there will always be a type I error when inferring a significant effect from ample to population. A percentage risk of 5% of type I error is acceptable (this is the alpha level). T tests are parametric analyses: data must be interval or ratio and there must be independence of observations (it must be independent measures). After experiment, the data must be normally distributed and there must be homogeneity of variance. Homogeneity of variance: The variance must be similar in all groups. You can test for this using Levene’s test, where we want NO SIGNIFICANT DIFFERENCE in variance between groups (showing that the variance is similar). If Levene’s is not significant (p > .05), then the assumption of homogeneity has been met. Look at the ‘equal variances assumed’ row. Reporting t test findings in APA format: T (df) = t statistic, p value, and explain what it means (there is a significant difference between males and females for enjoying Eastenders. Interpreting the t test: Look at the descriptive stats to indicate directions e.g. females had significantly higher enjoyment scores (M = 6.8, SD = 1.5) than males (M = 4.2, SD = 2.1). Create an appropriate graph to show significant findings (to 1 SD). Repeated measures t test: You look at sphericity and not homogeneity of variance. Sphericity checks whether each participant’s change in scores is consistent. You can test for this using Mauchly’s test. If Mauchly’s is not significant (P > .05), then the assumption of sphericity has been met. If it is significant and the assumption has been violated, then SPSS provides a “Greenhouse-Geisser” correction, where the degrees of freedom are reduced, meaning that a bigger effect needs to be found for a significant finding to be found. Results for a repeated measures t test: Was there a significant difference between groups? Use descriptive stats to point out the direction.