Wk 6 - ANOVA 1

Question 1

Why do ANOVA rather than t-test? (x1)

Answer 1

It guards against high family-wise error that would come from t-test error of .05 x number of comparisons

Question 2

How does ANOVA differ from t-test?

Answer 2

It tests multiple means against each other, while t-test can only compare one against the population

Question 3

How is ANOVA similar to t-test?

Answer 3

Both test the null hypothesis that there is no diff in means

Still comparing observed diff to expected

Question 4

What is partitioning in one-way ANOVA?

What are the names for two the portions?

Answer 4

Dividing the variance into components:
Between groups - treatment, individual diff, experimental error
Within groups - can only be ind diff and error
MStreat and MSerror

Question 5

Under what conditions will MSerror = MStreatment? (x1)

Answer 5

When there has been no effect of treatment

Question 6

Under what conditions will MSerror not = MStreatment? (x1)

Answer 6

When there has been an effect of treatment

Question 7

How does MS relate to variance?

Answer 7

It is the SS divided by df

Question 8

F-test has 2 types of df - why?

And how are they calculated?

Answer 8

You need one for the treatment, and one for the error
k - 1, and
N - k

Question 9

What does omnibus mean (re ANOVA)?

Answer 9

That we are looking at ALL the variance to see if the DV varies among IV levels

Question 10

What are the steps for a one-way independent groups ANOVA?

Answer 10

State hyps and IVs/DVs
SStotal = sum of all (scores minus the grand mean) squareds
SStreat = mean for each group - grand mean, squared, times number in that group, then sum all groups
SSerror = total - treat
df: total = N - 1, treat = k - 1, error - N - k
Find MS by dividing all SS by their df
Look up critical F for (treat, error)