Examine baseline equivalency. Here is one source: Anderson, M. A., & Maxwell, N. (2018). Baseline Equivalence: What it is and Why it is Needed. Submitted to AmeriCorps by Mathematica. Chicago, IL, September.

Yes, you can.

who are your respondents and why are they your chosen respondents? Jayashree Mahanti

Yes, you can especially as it is provided in the study: variables and sample

No, as scores on the pretest is only one variable that may account for differences in the two groups other than experimental treatment.

It will be better to assign at random as that would ensure equivalence on other variables that affect achievement such as motivation, grit, socioeconomic status etc.

نعم تصلح بعد اختيار العينة حسب السن ومستوى الدراسي

I can rely on the ratings from a pretest to apply a strategy or tool and then do a posttest to find out the effects of the strategy or tool. To choose the control group and the experimental group, you can base on the pretest results, as long as the two groups have similar characteristics of being affected by the problem

There are reports from different authors that support the use of quantitative pretest results to choose the control group.

Hi. I'm not really clear about the strategy. If you are establishing two groups with the same acquired knowledge, I do not see why one group is control and the other experimental. In principle, the control and the experimental groups must not have the same aptitude level otherwise the impact of your tool is nullified. Since your test subjects have the same knowledge at the start, you might as well use all of them as experimental group with their performance in the pre-test being the control reference point. You would have a more representative performance if you want to establish inter- and intra-subject variability. If all subjects in the pre-test are experimental, you would do well to test the same tool on candidates that you have not trained at all as control. Any other procedure might not produce a valid psychometric scale.

Others are plain wrong. You can and must match, if possible, the control and experimental group where both are equal. Since equal is impossible, the groups must be similar in most all respects. Here is a simple way to understand: You want to see if your intervention will lead to more students scoring >90% on a posttest. You pretest and sort control group A with an average of 37% and experimental group B with an average of 88%. At the end of the experiment, you find with your intervention group B did much better. The example simply explains why matching, which is more than just the pretest and should include other variables, needs a baseline equivalence.

Dear David.

Oh! Yes! All the others are plain wrong. But you are not asking him to do what he proposed to do. Did you really try to understand his strategy?

Kind regards

Dear Jimmy:

The poster stated the following: "I want to know if it is okay to use pretest scores to divide the control and experimental groups."

This action is classic A-B design, where one must consider baseline equivalence. I also stated other variables, such as demographics and design matter. Are you suggesting my example is wrong? Thank you.

David

Please see the classic book at the end. A-B design is common across all disciplines. There are weaknesses, and considerations of baseline equivalence, confounding variables, and lurking variables are a must. This book is one of many; any basic introductory psychology research textbook will show what I speak. Me? I am not a fan of A-B design, as there are many far stronger methods, but time, money, and ability continue to make A-B ubiquitous in education and psychology. If you conduct A-B, consider multiple factors.

Campbell, D. T., & Stanley, J. C. (2015). Experimental and quasi-experimental designs for research. Ravenio books.

Dear David,

Thank you for your observations. I do not think that we need text books for this particular case. I am making this reply purely because I'd like the poster to re-examine his strategy.

I for one endorse the principle to split his pretest subjects into two groups, but did you note what he wanted to do? He said: "Means If two students scores 30 in the pretest, I place one of them in the control group and the other in the experimental group?" In other words, he would end up with two groups having identical composition. Maybe you will forgive me if I've made a mistake, but please let me explain myself. Suppose I am running an experiment in musical pitch discrimination. I have two musically naïve subjects and two absolute pitch subjects. Now, I put one naïve subject and one absolute pitch subject in the control group, then I put the other naïve subject and the other absolute pitch subject in the experimental group. If now I run my experiment, I do not see what useful conclusion I would draw from the fact that I had two groups since the two groups are the same. If I run different experiments with the two groups, I cannot transpose the results since the other group did not undergo the test, etc.

Therefore, all of us who are plain wrong were possibly thinking of the usefulness of the control and experimental group having identical structure. It has nothing to do with right/wrong to use the pretest subjects. Once again, I am plain wrong, but for me, that is the end of the matter.

Have a great evening, David.

In my opinion yes, you can decide which experimental group is the one that obtains lower arithmetic mean scores, as some researchers have said