I have an experiment with 3 within-subject conditions: A, B and C. My dependent variable is reaction time. I am interested in comparing A to B and A to C. Using aov() on R I run an one way ANOVA with the two contrasts built in the model. Condition was a factor (in the sense of factor() ).

model1 = aov(reaction_time ~ condition + Error(subject/condition),
              data = My_data)
summary(model1, split=list(type = list("A vs  B"=1, "A vs  C" = 2)))

The summary function revealed that the overall model was significant. The contrast comparing A to B was significant (.012). The pvalue of the second contrast (A to C) was .0509. Since I have strong theoretical background for my comparisons, I divided the pvalue by 2.

One reviewer did not accept that. He claimed that since I reported a F test for the contrast (which is what aov() reports), I cannot conduct a one tailed test. I claimed that a F test comparing two groups is equivalent to a t-test comparing the same two groups. The obtained F value would the square of the t value, but the pvalue should be the same. Therefore, if I could divide the pvalue fo the t-test, I could similarly divide the pvalue of the F test.

What is wrong with my logic?


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.