6 - Comparing more than 2 samples

Question 1

What type of test is the ANOVA? must the data be normal?

Answer 1

It’s a parametric test and therefore data must be normal

Question 2

How can a tail of high values be normalised?

Answer 2

Log transform:
x = log10 (x)
If any values are 0, must use log10 (x+1)

Question 3

Why use variance? (4)

What is the convenient form?

Answer 3

Variances add up
We can partition variance between sources
We can calculate variance due to factor of interest
We can test if variance is more than expected
The convenient form is “sum of squares” (SS)

Question 4

What are the sums of squares?

Answer 4

The sum of the squared deviations from the mean

Question 5

What is the mean square?

Answer 5

the sum of squares
divided by the degrees of freedom
it is a standardised measure of variation

Question 6

What is the F statistic?

Answer 6

it tests whether the variation between samples greater than we would expect from the variance within samples?
it tests the ratio of two variances
if the two variances are equal, F =1 
the two variances each have separate df
F2,16 = 10.56 P = 0.001

Question 7

What test can you use to test for differences between samples in a one-way ANOVA?

Answer 7

Tukey test