Do the distributions of the two groups (experimental, control) have the same shape?

It is almost never a good idea to categorize continuous (interval-scale) data because you loose information (and statistical power) by "transforming" interval data into ordinal data. Have you examined the distribution shape in each group (e.g., by looking at histograms and boxplots) as well as compared the group variances? Also, t tests are fairly robust against violations of normality (and homoscedasticity as long as group sizes are about equal).

Christian Geiser These are not continuous data. They are counts.

@Blaine Tomkins and @Christian Geiser, thank you very much for the answers.

@Blaine Tomkins, by shape, do we mean the number of participants? If so, they are not equal size, but the proportion is at a reasonable level.

Hello Yusuf Polat. One way to address Blaine Tomkins' question about the shapes of the distributions would be to create a pair of stacked histograms, like this:

GRAPH

/HISTOGRAM=DV

/PANEL ROWVAR=GroupVar ROWOP=CROSS.

I've attached an example where DV=Age and GroupVar=sex to give you an idea of what it looks like. In this example, the shapes of the two distributions are quite similar. If you make a graph like this for your DV and GroupVar and upload it, readers may have a better idea about how to guide you. HTH.

Bruce Weaver, thank you so much for this detailed reply. I am sharing a histogram to give you an idea about my dataset. However, I am not sure how this will help in my case because my sample sizes are not equal, that is to say, in my Experiment group there are 46 students, but in Control group 29 students, and also I am looking for whether there is a significance between the login attempts of the groups rather than the shapes of distributions. I am not sure if I would really need visualization in the form of histograms.

Thank you very much once again, I really appreciate it.

ps: I just noticed that the the names of the previous people who replied to this post are not highlighted in my reply to them since I wrote my reply in my smartphone (not sure if it has got anything to do with it, might be related with my lack of RG knowledge :). Long story short, let me take this chance to mention them one more time.

Thank you so much Blaine Tomkins and Christian Geiser for taking time for this. 🙏🏻

Thanks for uploading that plot. The distribution shapes are fairly similar, but there is a clear outlier in the Control group. I think I would try a count regression model, likely a negative Binomial model.

I don't remember reading anything on the effect of outliers on count regression models, but I did just now find this simulation study, which may be helpful to you:

http://www.scik.org/index.php/cmbn/article/viewFile/5658/2980

I've only scanned parts of it quickly, but noticed that it mentioned quantile regression as another option "when we simply cannot obtain a reasonably fitted Poisson, negative binomial, and zero count models, because mean regression models may be sensitive to response outliers and provide no information on factors affecting other distributional points (e.g., upper and lower 5% quantiles) of the response due to some data problems."

HTH.

Bruce Weaver, thank you very much. I really appreciate it. I'll be sure look into what you have suggested. 🙏🏻