Bivariate Analysis: Chi-Square, Fisher, and McNemar Tests Flashcards
What is a Chi-Square Test?
Nominal or Dichotomous Variable, compared across 2 discrete independent groups
HOMOGENEITY
What is a Fisher’s Exact Test?
Dichotomous variable, compared two discrete independent groups
What is a McNemar Test?
Dichotomous variable, compared across Matched Measurements in two discrete dependent groups
PRE/POST
What is the Chi-Square Test of Homogeneity
Applied to a nominal variable from two or more different groups/populations to determine whether frequency counts are distributed identically across the groups
If a nominal variable is dichotomous, it means what?
There are ONLY two options
ex. improved vs not improved
For the Chi-Square Test of Homogeneity, the H0 compared what?
H0 = Pt = Pc
Proportion of patients with specific outcome will be EQUAL between treatment and comparison group
COMPARING PROPORTIONS NOT means
A large X^2 value in a Chi-Square tests = statistically significant, what can increase the value of x^2?
The larger the difference in proportions = greater x^2
A Chi-Square test is ONE directional with an alpha of 0.05 and a df =1, what is the the critical value?
3.84
What are the requirements/details of the Fisher’s Exact Test?
- More conservative than x^2, less chance of a false positive
- For 2x2 tables ONLY
- N<20 or if there are <5 in one of the cells
What is the Chi-Square Test of Independence?
Applied to 2 nominal variables from one group/population to determine whether frequency counts are distributed in a manner that shows an association between the two variables
What does Independence mean?
One variable does not say anything about another variable
What is the H0 and HA of the Chi-Square Independence Test?
H0: No association
HA: Association
What are the parameters of the Chi Square Independence Test?
- ONE Population
- TWO Dichotomous Variables
- FREQUENCY = ASSOCIATION
What are the requirements/details of the McNemar Test?
- Dependent groups, paired data
- Dichotomous variables, 2x2
- Compare proportions between PAIRED data