# Statistical test for replicate measurements with mean and standard deviation

I am having some trouble trying to find the right statistical test for my experiment. It is a flow cytometry experiment with many replicates and I am comparing normal and disease states. I get a median and rSD (or mean and SD) for each replicate, since I am looking at a population for each measurement. I can just use the means and use a t-test, but I would like to retain the spread (SD) data in the test. I'd appreciate any advice, I've searched all over before posting.

.....

(I had a feeling I was posting wrong)

The data look something like this, although with more replicates (the data are not paired):

+--------+---------+--------+------+
| Sample |  State  | Median | rSD  |
+--------+---------+--------+------+
|      1 | Healthy |  13471 | 6203 |
|      2 | Disease |   5521 | 1022 |
|      3 | Healthy |  13800 | 5299 |
|      4 | Disease |   4932 |  860 |
+--------+---------+--------+------+


The objective is to test whether the two groups (healthy and disease, in this case) differ significantly with regard to this parameter.

Just to get a sense of the data I am working with, see the example below. It is from a different experiment, but illustrates where the values come from nonetheless.

The median and rSD (robust standard deviation) are the standard measures of central tendency and spread for this kind of data, because of frequent outliers and non-normal distribution. I can also display the number of events, mean and SD with that software. Although I would much prefer to work with summary statistics, I can also export the listmode data, which would be a CSV file with the values for each individual event.

The "replicates" (may be the wrong word) are actually different samples from different specimens belonging to one of two groups (healthy/disease). Each sample has a population (of cells) that is positive for a certain parameter, similar to the population within the P2 gate of the image above. The idea is to measure that parameter for a certain number of samples for each group (about 10 - 20 each) and to determine whether the two groups differ significantly with regard to this parameter.

Does this help clarify?

Thank you again! Joe

• Replicates or a repeated-measures design requires a different test. Tell us more about your data and objective. – user2974951 Feb 12 at 14:31
• Please see below. I couldn't format my response properly as a comment. – Joefaa Feb 12 at 19:09
• You made a rookie mistake, you posted an update as an "answer". You should delete that post and edit your ogirinal question with the new information. You can do that by clicking the "edit" button below the tags. – user2974951 Feb 13 at 7:40
• As far as I understand, the median values that you have are actually aggregates? That is you summarized multiple observations to obtain that? I do not really understand what rSD is though, why do you report the SD of the mean if you have medians in your results? Also where are the replicates? Replicates imply multiple measurements of the same observation / sample. Do you have all the orignal data, non-aggregated? – user2974951 Feb 13 at 7:42