Chi Square Flashcards
What is the symbol for Chi Square?
χ2
When do we use Chi Square test?
When we have nominal data :) e.g. categories like male, female.
Chi square is usually used where we have __ variables, each of which are assessed on a _____ scale.
2
nominal
Do we use Chi Square on:
repeated measures design
OR
independent groups design?
Independent Groups Design
How do you report a Chi Square?
χ2 (df, n =….) = chi square, p = xxx
If the Chi Square has a significant effect what do we report?
The sample size e.g. n = xxx.
What does Chi Square Test?
The association between 2 nominal/categorical variables.
What is the descriptive statistic for a Chi Square?
A contingency table
How do you calculate the expected frequencies for each cell?
Multiply to row total by the column total.
Then divide by the number of participants.
Why do you calculate expected values for each of the cells?
To compare these with our observed values.
How is chi square calculated?
the sum of:
(o - e)2/ e
o = observed values e = expected values
Th chi square value will get ____ as the difference between the observed and expected values increases.
bigger
What is the standard way to report Chi Square?
χ2 (df, n = 100) = pearson Chi square, p = .xxx (two-tailed) [plus effect size if significant].
If the Chi Square is significant what do you need to do?
You need to explain what the association was using the observed and expected frequencies e.g. “more males (13) were smokers than we’d expect by chance (10)”
What do we need to double check for in any Chi square?
Expected cell counts!
Need to be less than 20% of counts under 5.
For chi square, you are only allowed to have fewer than ____ of your expected cell counts under 5.
20%
What do you do if you have more than 20% of your expected counts under 5?
Report the Fisher’s Exact Test.
just report the p value
What should you use for your chi square effect size? If you have a 2 x 2 table?
Phi Coefficient
What effect size is used for more complex data?
Cramer’s v
What does the phi coefficient range from?
0 to 1
A value closer to 1 reflects what?
A larger effect!
What is a small effect size?
.1 to .3
What is a medium effect size?
.3-.5
What is a large effect size?
over .5
What are the 2 ways to deal with low expected cell counts (e.g. more than 20% under 5)
- you can ignore a group
2. you can merge 2 groups together
For G power:
What proportions do we have to set?
set proportion p 2
people in treatment group/outcome
set proportion p1
people in control group/outcome
For G power:
how do we change the allocation ratio?
If there is 50 people in each group then fab it’s 1!
If not divide smallest group/ bigger group and get a number :)
Chi square is a ____ test.
non-parametric
