If you are familiar with R, try  the 'ez' package, as it allows mixed ANOVAs as the one you are describing (within and between variables). You can see how with '?ezANOVA'.

I hope this can help you.

One option would be a form of logistic regression that controls for the correlated nature of the repeated measures.  The two most common ways to do that are generalized estimating equations (GEE) and multilevel modeling.  Perhaps someone can give more detailed advice if you tell us what software you use. 

HTH.

Thanks Bruce. Shall I use the logistic model in GEE. And is the data entry different for GEE or is the data entered the same way as in a mixed ANOVA?

How the data need to be arranged may depend on what software you are using.  In SPSS, for example, you need a LONG file, with one row per observation of the outcome variable.  Therefore, you need an ID variable to indicate who each data point belongs to, and variables indicating which time point it is, etc.  HTH.

Thanks a lot Bruce. I am using SPSS. I think GEE is what I need for my analysis. 

Hi Inaamul.  I estimated a model something like this a few years ago.  For reasons I can't remember, I created my own indicator variables for the 3 time points, and included two of them in my model.  In hindsight, I could have included the original Time variable (with values 1, 2 and 3) as a "factor".  Here is the syntax I used (with 2 indicators for the 3 time points) plus what I think it would look like with Time as a categorical variable.  HTH.

****************************************************************** .

*** GEE: Quit = const + treat + t2 + t3 + t2*treat + t3*treat' *** .

****************************************************************** .

* mrnum = medical record number = ID variable.

* Generalized Estimating Equations.

GENLIN quit (REFERENCE=FIRST) WITH treat t2 t3

/MODEL treat t2 t3 treat*t2 treat*t3 INTERCEPT=YES

DISTRIBUTION=BINOMIAL LINK=LOGIT

/CRITERIA METHOD=FISHER(1) SCALE=1 MAXITERATIONS=100 MAXSTEPHALVING=5 PCONVERGE=1E-006(ABSOLUTE)

SINGULAR=1E-012 ANALYSISTYPE=3(WALD) CILEVEL=95 LIKELIHOOD=FULL

/REPEATED SUBJECT=mrnum WITHINSUBJECT=t2*t3 SORT=YES CORRTYPE=unstructured ADJUSTCORR=YES

COVB=ROBUST MAXITERATIONS=100 PCONVERGE=1e-006(ABSOLUTE) UPDATECORR=1

/MISSING CLASSMISSING=EXCLUDE

/PRINT CPS DESCRIPTIVES MODELINFO FIT SUMMARY SOLUTION (EXPONENTIATED).

* Instead of creating indicators for time, I could have treated

* the original time varialbe as a "factor", something like this,

* assuming time has values of 1, 2 and 3 for 3 time points.

* The reference category for time might be different than in my

* model above, but that could be changed with appropriate syntax.

* Generalized Estimating Equations.

GENLIN quit (REFERENCE=FIRST) BY time WITH treat

/MODEL treat time treat*time INTERCEPT=YES

DISTRIBUTION=BINOMIAL LINK=LOGIT

/CRITERIA METHOD=FISHER(1) SCALE=1 MAXITERATIONS=100 MAXSTEPHALVING=5 PCONVERGE=1E-006(ABSOLUTE)

SINGULAR=1E-012 ANALYSISTYPE=3(WALD) CILEVEL=95 LIKELIHOOD=FULL

/REPEATED SUBJECT=mrnum WITHINSUBJECT=t2*t3 SORT=YES CORRTYPE=unstructured ADJUSTCORR=YES

COVB=ROBUST MAXITERATIONS=100 PCONVERGE=1e-006(ABSOLUTE) UPDATECORR=1

/MISSING CLASSMISSING=EXCLUDE

/PRINT CPS DESCRIPTIVES MODELINFO FIT SUMMARY SOLUTION (EXPONENTIATED).

Hello,

The GEE idea is good.

You may also try a mixed logistic model. In R, that would be: http://www.ats.ucla.edu/stat/r/dae/melogit.htm 

In your case, I would write something like that 

m

Agreed with Bruce Weaver and  Jean Paul Maalouf

Thanks all. Bruce, you have been great help. But I had to do some reading and video watching about GEE in SPSS before I could proceed the analysis. And now I'm ready. Thanks :)