Hello Elleonora,

Could you elaborate a bit on what "relationship" is being referred to? I will agree with you that the quotation from the anonymous reviewer doesn't make sense as stated; perhaps this person was thinking of something else. Nor does it necessarily sound like a mandate (for you to engage in further analysis), as opposed to an observation that perhaps another analytic approach might have been applied (which is often the case).

No matter; the primary issue concerns the research question(s) you were trying to address, and the resultant analyses and your interpretation thereof.

Good luck with your work.

Eleonora Marzilli , the term "confirmatory correlation analysis" is quite uncommon (and in some sense even contradictory by itself I would say); is used sometimes in medicine to refer to the correlation analysis -with the Pearson or non-parametric coefficients- that is performed after a t-test/(M)AN(C)OVA to "confirm" the direction and strength of the associations found by the t-test/(M)AN(C)OVA. See for example:

Article Activation/modulation of adaptive immunity emerges simultane...

Article Core temperature correlates with expression of selected stre...

Hi,

I agree with David Morse . I would send an email to the editor with the request to clarify what the reviewer meant. This is better than doing something that will not match the reviewer's request and postpone the whole process (or worse).

All the best,

--Holger

David Morse , thank you for your comment.

I carried out a series of MANCOVAs to verify if children's genotype (categorized in two groups) was associated to children's psychopathology and parental psychopathology. Thus, i conducted a series of MANCOVAs, with group as factor, the scores of other instruments assessing children (or parental) psychopathology as dependent variable, ad children's age, and sex as covariates. Then, i reported the univariate significant effects, discussing significant associations emerged.

I had a very small sample size, so I thought to conduct this kind of analysis to get preliminary results.

I could conduct Pearson correlations, separately in the two groups, to confirm the associations emerged between the variables, but I do not know if it makes sense.

Moreover, the same reviewer requires a posthoc test that allows consolidating differences of statistical significance. To my best knowledge, post-hoc test are necessary with more than two groups. I feel confused

This statement " I had a very small sample size, so I thought to conduct this kind of analysis to get preliminary results. " would bother me even more..... I think it is hard to draw valid conclusions from small samples, since they tend to show too large effect sizes and are not really suiteable as pre-studies (or what do you mean with preliminary results?). See for example Loken & Gelman (2017) or Schönbrodt & Perugini (2013)

Rainer Duesing , the sample is composed of 100 families.

I consider it small in order to carry out eventual mediation, that the conceptual model would have suggested. It is a pilot study. The reviewer suggested to implement the analysis to give more value to the results that emerged.