Hi Fahime,

The attached file lets you find the answer.

You have two choices:

1 - test the difference between two proportions

2 - The difference between the frequencies of answers using the chi-square test

I'm not sure I understand your study design. Here's what I think you mean:

1. Each participant views several trials.

2. There are two conditions, experimental and control (E and C).

3. For condition E, the expected (correct?) answer is Yes.

4. You've not said what the expected (correct?) answer is for condition C--is it No?

If all of this is correct, I would suggest that you consider using the GENLIN procedure to fit a binomial logistic regression model using generalized estimating equations (GEE) to account for the correlated nature of the repeated measures data (i.e., each participant has data for both E and C conditions). There are various ways to set it up that would yield the same results (I think). I think the way I would do it is have a data file with 2 rows per person, one row for each condition (E and C). And the variables would be:

Exp (1 for condition E, 0 for condition C)

NumTrials -- the number of trials in the condition on that row

NumYes - the number of YES responses

Then, for GENLIN, use the "events-var OF trials-var" formulation, like this:

GENLIN NumYes OF NumTrials BY Exp...etc

To get into the dialogs that allow you to use GEE to deal with the repeated measures data, click on Analyze > Generalized Linear Models > Generalized Estimating Equations. (You may be able to find an example in the Help files.)

With this type of model, the coefficient for Exp will give you the difference in the log-odds of YES responses for E vs C. And the Exp(B) column will give you the odds ratio. B=0 and Exp(B) = 1 would indicate that the odds of YES are the same in the two conditions.

I hope this helps.

Your data is a categorical data (considering, yes=1, no=0). You can use Chi-square test for that.

To respondents who are suggesting that Fahime use "a chi-square test", notice that the study design appears to entail having multiple Yes/No trials per person in two different experimental conditions. So you need to clarify how the chi-square tests you're proposing take into account the correlated nature of the multiple trials within-subjects. The ordinary chi-square test of association will not work in this case, because it requires independence of observations (which typically translates to one observation per person).

This question is impossible to solve, unless you tell us with information do you have on other exposures and if a person is able to do the test repeated times.

It also depends on your hypothesis … if you really want help, tell us: what is you hypothesis and what else have you recorded to compare with...

Hey, we don't. bite ;)

All of the answers above may be right or wrong depending on what else do you have.