# Comparing withdrawal rates for a course

#### Vthompson

##### New Member
Hello!!
Which statical test should I use to compare withdrawal rates for a course for two semesters. I know I can report the means but need to know if the difference is significant.

For example:
Fall2017 10 out of 33 students withdrew
Spring2018 5 out of 66 students withdrew

#### staassis

##### Active Member
Z-test for two proportions.

#### Karabiner

##### TS Contributor
Alternatively, Chi² test of association (withdrawal [yes / no] crosstabulated with year [2017 / 2018]).

With kind regards

Karabiner

#### Vthompson

##### New Member
Excellent. Thank you for confirming what I thought it was.

#### hlsmith

##### Less is more. Stay pure. Stay poor.
I would recommend looking at rate differences, in that they will provide an estimate with confidence interval:

30.3% vs. 7.6%
So, 22.7% (95% CI: 5.8%, 40.0%),

Code:
/*SAS CODE*/
data TS;
input Exposure Outcome count;
datalines;
1 1 10
1 0 23
0 1 5
0 0 61
;

proc freq data=ts order=data;
table exposure * outcome / riskdiff;
weight count;
run;

#### ondansetron

##### TS Contributor
I'd probably also look for a CI method that preserves the nominal coverage. For a difference in two proportions, I believe the Agresti-Caffo method is a good option and if just trying to get a CI for each proportion, then the Agresti-Coull method is a good choice (but there are others with different performance characteristics).

#### Vthompson

##### New Member
Thanks every one. You’ve been such a great help!