1. ## test of proportion

I have forgotten my basic statistics...

I have a group of people who answered two different questions. I have been asked to do a test of proportion to see if the percent who answered agree on the two questions is statistically significant. That is to see if there is a statistically significant differences between the number agreeing with one question and agreeing with another.

I can do that of course, what I am not sure of is if you can do a test of proportion with the same population this way. All the examples I found are testing two different populations. Whether it makes sense to do a test at all when you have the whole population (which they do) is of course another question.

2. ## Re: test of proportion

Yup, chi-sq test given you don't have expected values in your 2x2 table </=5. So big picture you are looking to see if the proportion of people agreeing or disagreeing is the same for the other question.

If you think you have the end-all-be-all population, just don't look at the significance test and embrace the proportions.

P.S., you can also use the "/relrisk" on the table line to get an effectsize.

3. ## The Following User Says Thank You to hlsmith For This Useful Post:

noetsi (05-25-2017)

4. ## Re: test of proportion

I don't understand the point about a chi square. I want to do a test of proportions, are you saying that you can't with the same population or that a chi square test is better? All I have is the aggregate numbers, not individual responses.

Just because I have tens of thousands of cases and you have to expend vast efforts to get 100, many of whom drop out of the intervention at inconvenient times, is no reason to get uppity on me

I don't understand this at all.
P.S., you can also use the "/relrisk" on the table line to get an effectsize.

5. ## Re: test of proportion

Code:
``````data noetsi;
input q1 q2 count;
datalines;
1    1    560
1    0    250
0    1    200
0    0    800
;

proc freq data=noetsi order=data;
table Q1*q2 / chisq relrisk;
weight count;
run;``````
Reject null hypothesis of homogeneity of proportions and those answering agree to Q1 have 3.5 greater incident of answering agree to Q2 than those answering disagree to Q1.

 Tweet

#### Posting Permissions

• You may not post new threads
• You may not post replies
• You may not post attachments
• You may not edit your posts