0
$\begingroup$

I have two variables (group, achievement level). I want to check the effect of an intervention on different achievement levels. For example, I want to check whether there is an increase in achievement scores in low achievers, average and high achievers. Please suggest me which statistical tool should I use? Shall I use individual t-test as I have found in many articles? Or I should use mixed ANOVA?

$\endgroup$
  • $\begingroup$ You say "an increase in achievement scores", do you have achievement scores from both before & after the intervention? How did you arrive at the groupings? Did you simply categorize the participants based on prior achievement scores? $\endgroup$ – gung - Reinstate Monica Apr 15 '15 at 16:16
  • $\begingroup$ Yes I have both scores. I have categorized them on the basis of their summative scores. $\endgroup$ – kaushal Apr 15 '15 at 16:17
0
$\begingroup$

You say achievement level is an independent variable, then talk about it as a dependent variable. Do you have two scores per person (e.g. a pre-treatment score and post-treatment score) and you are interested in the difference between those scores? If so, it sounds like what you want is simple factorial ANOVA. Predict post-treatment achievement levels based on group, pre-treatment achievement levels, and their interaction. The interaction will tell you if the effect of treatment depends on the level of pre-treatment achievement level.

$\endgroup$
  • $\begingroup$ Dear Andrew,thanks for your early response. Yes, I do have both pre and post scores for each participant.I want to find difference between groups for all levels. $\endgroup$ – kaushal Apr 15 '15 at 16:11
  • $\begingroup$ So you want to find the average difference between groups? Its the same things, but leave out the interaction. $POST=GROUP+PRE$. Technically speaking, you can leave the interaction in there. If there is no interaction, then the models are equivalent. If there is, then that may be of interest (e.g. people responded, on average, to treatment (the main effect of GROUP), but people with low scores responded better than people with high scores(the interaction between group and pre-treatment scores)). $\endgroup$ – le_andrew Apr 15 '15 at 16:17

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.