One Way Anova

Question 1

T test vs regression

Answer 1

Diff between means vs modelling doff between groups (line of best fit. Gives same output but t test only for 2 means. Regression more so known as general linear model GLM

Question 2

Multiple comparisons

Answer 2

More groups means more comparisons, higher chance of type one error called familywise error. For 0.05, prob of type one error calc by 1-(0.95) to the lower off k (no comparisons). If correct alpha, miss real effects (type 2) so use and a.

Question 3

One way anova assumptions

Answer 3

One iv, but can have more than 2 levels, one dv, normality distribution, homo of variance. Data are independent. V robust so if violated use unless sig bad

Question 4

Reporting results

Answer 4

F value is like the T value, use p to see if sig. to write results do group size, sd/se, results of assumptions, f value, df for between and within (model and error). To report F(between df, within df)= f value, p=0.05.

Question 5

Theory behind it

Answer 5

F ratio is explain to unexplained variance. There’s a big formula but don’t need to know??

Question 6

Follow up tests

Answer 6

F ratio says main effect of group but doesn’t say where effect is coming from so need follow up, only run if sig. two kinds: planned contrasts/ comparisons used for testing a hyp, smaller no comparisons or post hoc: use if don’t know where diff coming from