My problem is I have 3 biological replications (reps) withe each having 4-5 technical reps with two of the biological reps having comparable results for the treatment and control.

These values represent the colonization of donor stem cells. (Treatment is about 50% of control) whereas in the third rep I still have ~50% drop from control but the values for both treatment and control are drastically reduced.

Is there a statistical method to analyze this data because using a t-test obviously results in a non- significant p value? Or would it be best to just do an additional rep and throw out the last rep as an outlier as results from other experiments are comparable to the last 2 reps?

Example - Biological reps below, average of technical reps
Rep 1 control 300 treatment 150
Rep 2 control 230 treatment 112
Rep 3 control 24 treatment 11


Two comments:

-There may be true biological variation and the data from rep 3 may be correct, and not some kind of technical mistake (which is why, I presume, you want to remove those values as outliers).

-Each biological rep has its own control. So tabulate the difference or better (in this case) ratios. Your data look incredibly consistent, with the treatment cutting the response in half (approximately).


Pair the data into vector of two elements and apply Mahalanobis distance to detect two dimensional outliers. If it is then remove it. However, in the example it seem pretty obvious. The theory of your discipline should say something about what is and what is not an outlier, though. So proceed with caution as always with statistical outliers.


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.