Levene's test for equal variances, Tukey's honest significant differences

Question 1

what do we do before using anova?

Answer 1

• construct boxplots to compare symmetyr and spread
• use a QQ plot to test for normality
• check equal variances - test null hypothesis that all populations have the same SD

Question 2

Levene’s test for equal variances

Answer 2

test statistic has an f distribution

1. for each value calculate the value of difference from each group mean
2. calculate the treatment sum of squares
3. calculate the mean sum of squares for treatments and for errors
4. calculate test statistic

Question 3

tukey’s honest significant differences

Answer 3

• preformed after ANOVA
• asks: which pairs of populations have different means (could do a 2-sample t-test but multiple testing is a problem)
• this corrects for multiple testing
• same assumptions as ANOVA

Question 4

how do you do tukey’s?

Answer 4

calculate the statistic for each pair of populations and work out if the differences between the means is significant