Hello!
I have averaged the duplicates of my raw Ct values and have used the delta delta Ct method to get a fold-change for each of the data within each subgroup. Each subgroup has around 4-6 lots of data which are averaged to give me one final average fold-change per subgroup.
I saw within one subgroup for example, that one of the fold changes was 0.55 whilst another was 3.55 and the other data between spanned from 0.6 to 1.0. So I removed the data which gave the 3.55 fold-change.
- Is removing data acceptable to do?
-Are there a minimum number of Ct values required to take an accurate average, or can just two values be adequate if they are producing fold-change data within close range of one another?
- When it comes to removing outliers, just how far out must a value be for it to be classed as skewing the data?
My issue is that due to the number of random outliers such as that 3.55 value as just an example, I have not had any significant results where there technically should be! And when I remove the outliers, I have significance in my results.
I have attached an example of what I mean, and have highlighted certain values that I would remove in red. I hope someone can let me know whether it would be acceptable for me, scientifically and ethically to do so!
Best wishes
Your question is not specific to qPCR - it is the general question waht to do with outliers. If a data value is "outlying", then this is first of all only an indication to have a closer look for possible reasons and a plausibility check. If the value is highly implausible (unphysiological, physically impossible), then it must be removed. Then you can check if everything is ok with the sample and the measurement (is there a note in the lab book, is the volume of the PCR ok? How does the amplification curve look like? etc). If you identify a problem here, the value must be removed. If everything is fine - just the value is a bit "off", then it is a value you must take as serious as all the other values!
You made the mistake to look at fold-changes, which have a strongly right-skewed distribution. You will always tend to see outliers when looking at fold-changes, but that is normal and these values shouldn't be removed. You must look at (an analyze!) the deltaCt values. I did that for the data you attached and I don't see any outliers at all. The distributions look similar in all subgroups (left plot) and the residuals (against the subgroup means) are in good agreement with the normal distribution (right plot).
I am not a statistician but looking at your excel sheet, I see one of the data-point ie H1E has a value which is far from rest other data-points of that particular group and could be treated as an outlier ( to make your data look neat). I usually run GraphPad Outlier calculator to correct for outliers.
- Badges
- Science method