Lecture 104 - Categorical Variables

Question 1

Degrees of Freedom for a Chi squared test =

Answer 1

Degrees of Freedom for a Chi squared test = (Rows - 1) x (Columns - 1) –> remember rows and columns do NOT included totals.

Question 2

Expected Value =

Answer 2

Expected Value = (Row total x Column total)

Grand total

Question 3

How does the comparison of P-value to alpha describe significance?

Answer 3

p-value > alpha –> no significant difference

p-value < alpha –> significant difference

Question 4

For which values of Expected values should you use a X² test, and for which should you use a Fisher’s Exact test?

Answer 4

For Expected values > 5 –> X²

For Expected values < 5 –> Fisher Exact test

Remember to use significance tests when Confidence Intervals overlap

Question 5

95% CI = ____ +/- _____ x (square root [(pq)/n])

Answer 5

95% CI = p +/- 1.96 x (square root [(pq)/n])

Question 6

When Expected Values get closer to 5, use ______ ______ test for most accurate results. It is possible that X² p-values for Expected values close to or < 5 will show significance when there actually isn’t.

Answer 6

Fisher’s Exact test

Question 7

What does an Odds Ratio < 1 tell us about the relation between the dependent and independent variables? What about OR > 1?

Answer 7

OR < 1 means the independent variable is Negatively correlated with the dependent variable (e.g. protective quality of a medication for a disease –> people who took the med (independent variable) were LESS likely (negatively correlated) to get a disease (dependent variable)).

Question 8

When looking at CI for OR, how can we determine if results are statistically significant?

From the attached image, which variables showed statistically significant results?

Answer 8

If the range of values within the CI for the OR does NOT include the value 1, the results are statistically significant.

Age, Physician, DOT, and Institutions showed statistically significant results (CI did not include 1)