2
$\begingroup$

I have measures from an epigenetic variable from two different groups (Treatment and control) measured in two different tissues. I have identified differences for this epigenetic variables between Treatment and control groups (Group) in each tissue separately, splitting my data into two subsets (one for each tissue) using the following models in R:

model_tissue_1 <- lm(Epigenetic_variable ~ Group + Age + Sex + Cell_proportion_1 + Cell_proportion_2 + Cell_proportion_3 + Cell_proportion_4, data = my_data_tissue_1)

model_tissue_2 <- lm(Epigenetic_variable ~ Group + Age + Sex + **Cell_proportion_1 + Cell_proportion_2 + Cell_proportion_3 + Cell_proportion_4 + Cell_proportion_5 + Cell_proportion_6** + pH, data = my_data_tissue_2)

Please note that each model includes different cell proportions as covariates, according to the studied tissue. Also, model_tissue_2 includes an additional covariate that was not measured in the other tissue (pH).

Now, I would like to evaluate cross-tissue differences for the Epigenetic_variable among the groups, i.e., to identify Epigenetic variable differences between tissue_1 and tissue_2 for individuals in the Control group, as well as in the Treatment group, if possible considering all the covariates from each model. Is there a way to do this?

Thanks :)

Improve this question
$\endgroup$
5
  • 1
    $\begingroup$ Why did you split the data rather than include tissue as a covariate ? $\endgroup$
    – Robert Long
    Commented Jul 30, 2020 at 4:09
  • $\begingroup$ Because I wanted to explore the differences on Epigenetic_variable in each tissue and to consider the different cell proportions from each tissue, including them as covariates. $\endgroup$
    – Brenda
    Commented Jul 30, 2020 at 4:43
  • 1
    $\begingroup$ That doesn't tell me that you should split the data. Unless I am missing something, splitting the data means you can't answer your research question. $\endgroup$
    – Robert Long
    Commented Jul 30, 2020 at 5:52
  • $\begingroup$ Thanks for your reply. I will merge my data and make new groups considering both the experimental group and tissue, i.e., control_tissue_1, control_tissue_2, treatment_tissue1 and treatment_tissue_2 and include the cell proportions and other variables from both tissues as covariates. Then, I will performe comparisons among the groups. Do you think that this approach will be ok? $\endgroup$
    – Brenda
    Commented Jul 30, 2020 at 16:22
  • $\begingroup$ Yes, that would be my advice. $\endgroup$
    – Robert Long
    Commented Jul 30, 2020 at 16:42

1 Answer 1

Reset to default
2
$\begingroup$

I am not 100% clear about this but my understanding is that a dataset has been split by tissue type. A model has been fitted on each subset, using largely the same variables in each model (at least one variable is constant/absent for one tissue type).

The research question centres on a comparison of the estimates for some of the variables between the two models.

If I have understood this correctly, then I don't think the estimates can be compared in a meaningful way. The way I would approach this is to work with the whole dataset and include a variable to identify the tissue type. This may need to be interacted with other variables to answer the research question.

Improve this answer
$\endgroup$
2
  • $\begingroup$ Yes, the dataset was split by tissue type into two subset, and a model was fitted for each subset with same variables in each model. Also, I included specific covariates for each model that are not present in the model (cell proportion from each tissue and pH). Thanks for your recommendation, I will test interactions among the variables in a model fitted for the whole dataset (both tissues). $\endgroup$
    – Brenda
    Commented Jul 30, 2020 at 18:53
  • $\begingroup$ Does this answer your question ? If so, please consider marking it as the accepted answer, and if not please let us know why. $\endgroup$
    – Robert Long
    Commented Aug 21, 2020 at 12:11

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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