Methodspace - home of the Research Methods community

I need your help for quasi-experimental study and analysis  so Please reply me for below noted questions.

 

1. Can  t -test for control and experimental groups ( Non-equivalent quasi-experimental ) for pretest data  and t-test for post test data sufficient to show the effectiveness of the implemented program? Because levene's test shows violation of assumption for univariate analysis of the data.

 

2. Is it fine if one uses Welsh's t-test though normality assumption violated? The groups are not same in size.

 

Please let me know.

 

Your response is highly appreciable.

 

thanks.

Tags: ancova, and, quasi-experimental

Views: 657

Reply to This

Replies to This Discussion

Hello, Dear Dave

 

It's quasi-experimental non-equivalent design. It was already carried out, now the problem is how to infer and reach with strong causality? Because, ANCOVA assumptions violated then what to do? Let me know the book which deals with Johnson -neyman approach and rank ancova also.

Or should one should go for mann-whitney tests on both the pretests and post tests as well? and then pre-to post should be dealt with wilcoxin matched pair sign tests?

But no one can replace ancova. Right?

Let me know the effect of non-normality over welch's t- test also.

thanks.

No No No.

Please cite any quasi-experimental NON-EQUIVALENT CONTROL GROUP study which graph the change and that did not use t-test or ancova or any other statistical technique.

In such study, why do we take control group?

the rationale for taking control group is to compare the data statistically.

When groups are assigned randomly to either control or experimental, then also one can go for ANCOVA and t-test. The Donald Ary, Mitchell, Gay and ariasan and many other have mentioned parametric statistical analysis.

Let me know your reply.

Thanks

RSS

Follow us:

© 2014   Created by SAGE Publications.

Badges  |  Report an Issue  |  Terms of Service