0
$\begingroup$

I would like to explain one dependent (scale) variable with three independent variables (all categorical). As far as I understood I can try Factorial Anova. Is it true? Also, my dependent variable may affect the combined effect of these three independent groups. By using this method, can I measure it?

my dependent variable is performance and I try to understand which factor affects the performance of people more. As I had three different conditions, I have results from the same people more than once. And in one point, I used the effect of two conditions together to see if this will make things worse. You can see the example here.

enter image description here

Thanks

$\endgroup$
  • 1
    $\begingroup$ We would need more information about your variables to better help. What are the variables? How many levels in each independent variable? Was the design within-subjects, between-subjects, or a combination (e.g. did you measure more than once from an individual participant or subject? If so, which variable accounts for the multiple testing points in time or "carries" this timing information in some way). Do you have several missing values or many more outliers in your DV than one would expect? What type of data is your DV (e.g. count data of individual behaviors, a continuous measure like body $\endgroup$ – Thomas Wukitsch Aug 8 '19 at 17:09
  • $\begingroup$ @ThomasWukitsch We would not normally pile so many questions on to a new user at the same time. $\endgroup$ – James Phillips Aug 8 '19 at 17:21
  • 1
    $\begingroup$ @JamesPhillips 5 direct questions with some examples tossed in is too much? I would need to ask all of these anyway to answer the question. If i ask them one at a time it is going to take much longer for them to get an answer. I do this with my undergraduates who have stats questions pretty frequently. $\endgroup$ – Thomas Wukitsch Aug 8 '19 at 17:27
  • $\begingroup$ @ThomasWukitsch I understand your justification for this behavior. $\endgroup$ – James Phillips Aug 8 '19 at 18:53
0
$\begingroup$

If you do not have any other independent variables (e.g. age controls or something like that) and you only have a limited number of unique treatments (e.g. Condition A, B, C or AB BC etc.) it seems most logical to 'just' compare mean performance per group. If there aren't any other shared regressors, there isn't much reason to complicate matters: a model with only a number of unique categories will simply yield the mean values of each group. You can then do standard testing on the means per group.

I am just judging from the screenshot you showed - maybe you have a more complicated setup which justifies additional modelling.

| cite | improve this answer | |
$\endgroup$
  • $\begingroup$ Thanks for the comment @MarkVerhagen , I checked the mean values and they really explain the main idea. But I will use this for my thesis and my advisor told me that it would be nice to show the significance of findings as well. This is why I was thinking of a model to run. $\endgroup$ – user255928 Aug 9 '19 at 12:36
  • $\begingroup$ Just creating separate regressors for each unique group with binary values if the person was or wasn't part of that group and omitting one (as the comparison) from the regression is a way to do this. The coefficients will indicate whether the groups are statistically different from the omitted group. If you have a logical control group, use that one and you can just evaluate the significance. $\endgroup$ – Mark Verhagen Aug 9 '19 at 15:00

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.