I ran a two-way between subjects ANCOVA with drug (2 levels-drug and no drug) and groups (depressed, confused and normal) as the IVs and the number of errors on a cognitive task as the DV. I included age as the covariate in my ANCOVA and it turns out to be highly significant (p < 0.001) and has the largest effect size. Groups were not age matched in any way, samples were equal across the three groups and 2 drug conditions, the design is between subjects and all assumptions for ANCOVA were met. I was wondering how I can follow up this significant effect of the covariate. I ran a correlation and it turns out the covariate is highly significant with the DV. What can I do to further explore the relationship between the two?

  • $\begingroup$ Could you explain the context of the covariate? Is it a variable you captured as background information or is it an variable that you controlled? That makes a big difference in how you handle it. $\endgroup$ – ReneBt Mar 16 '18 at 11:11
  • $\begingroup$ Hello and thank you for your quick response! I have two IVs which are group and drug, one DV which is the numbers or erros participants make on a task and the covariate is age. It is a variable that I controlled for in my ANCOVA and it comes up with a highly significant main effect. $\endgroup$ – KSI Mar 16 '18 at 11:56
  • $\begingroup$ I have a lot more questions about the study design which would help answer your question. Could you update your question to provide a lot more detail. Were samples age matched in anyway during assignment to group and drug category. Could you give more insight into your IVs, how many strata do they create together and how are samples distributed across them (equal, vastly different or somewhere in between)? How are your ages distributed? $\endgroup$ – ReneBt Mar 16 '18 at 16:39
  • $\begingroup$ No, samples weren't age matched during assignment to group and drug category. My two IVs are drug (2 levels-drug and no-drug condition) and group (3 groups-depressed, confused, normal). Samples are equal across the 3 groups and 2 drug conditions. Plus all the assumptions for ANCOVA were met. $\endgroup$ – KSI Mar 17 '18 at 10:33
  • $\begingroup$ OK, this is useful information, but leads to more questions. So you have one IV (drug condition) based on (random?) assignment and one IV (group) based on (random?) selection from a larger pool for each underlying condition, is this correct? Is the data age matched across drug within each group? Could you please update your question with as detailed a description of you design as possible so that any answer can be appropriate to the design? $\endgroup$ – ReneBt Mar 19 '18 at 13:21

Your Answer

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

Browse other questions tagged or ask your own question.