week 8 Flashcards
which test assesses change over two time points on a single dichotomous variable
McNemar test - change between two time points ex. students thoughts on if a class should be required before and after taking the class
cells B and C in a mcnemar test represent
change
Mcnemar test only works for
dichotomous variables
critical values of chi are always looked up with an
unsplit alpha
chi square is an
omnibus test - shows that cells differ but not which ones
how do you calculate DF for a mcnemar test
DF=1
How are standardized residuals calculated in a bivariate chi-square analysis?
By dividing the difference between observed and expected frequencies by the square root of the expected frequency.
When will the phi coefficient and Cramer’s v be identical?
When at least one of the variables has only two categories.
When calculating Cramer’s v, what does k represent?
The number of rows or columns in the cross tabulation, whichever is less
phi coefficient always ranges from
0-1
t/f you can not have a negative phi coefficient
true because it is directly estimated from chi-square which is always positive
as the size of cross-tabulation increases so does
value of chi-square
how do you calculate the phi coefficient?
square root of obtained value of chi-square divided by overall N size
how to calculate expected value
product of the marginal totals for the row and column in which the cell is located
the simplest way to conceptualize a difference between two time points
mean difference
what is the main difference between univariate and bivariate chi-square analysis
expected value calculations
what is the bivariate chi-square
association between two categorical variables
two bivariate applications of chi square
- Change over time within a single variable
- McNemar test - Association between categorical variables
- phi coefficient, cramer’s v
what kind of statistic is the mcnemar test
chi square statistic
what kind of analysis is the mcnemar test
A repeated measures analysis
- must survey the same group of individuals twice
- intended to evaluate change over time
what is null hypothesis for mcnemar test
no change
in a 2x2 chi square analysis what will the DF be
always 1
the mcnemar test only works for
dichotomous variables and designs with only two time periods
cramers v and phi coefficient test what kind of data
nominal
in a cross table what are we comparing
model fit for our observed data
what can standardized residuals be treated as
may be treated like z-scores, with scores greater than +2 or less then -2 considered to be significant
when can we consider a cell to be significantly different from our expectation
If the standardized residual is greater than +2 or less then -2
t/f: chi square helps determine the magnitude of association
FALSE- it identifies if there is an association but not the magnitude
phi coefficient ranges from
0-1, cant have negative phi coefficient bc it is a direct estimate from chi square and chi square cant be negative
What can larger cross tabulation tables result in?
Phi coefficients > 1
- as the size of our cross tabulation (ie as the number of cells increases) the value of chi square tends to get larger. this is why the df for chi-square analysis is based on the number of rows and columns in our cross table
- this systematic increase in the size of the chi-square makes it increasingly likely that the chi coefficient will blow past our conceptual upper limit of 1 - thus we need to apply a correction factor to our phi coefficient so that the correlation still ranges from 0-1
cross tables larger than 2x2 require a
correction
what is the simplist corrected cross-tabular association
cramers v
rule for chi square
80% of expected values have to be greater than 5, all expected values must be larger than one
the magnitude of association is identified using
a phi coefficient or cramers v
