I am comparing cell staining with vs. without treatment. The treatment certainly has an effect and results in unmistakably brighter staining. I quantify >200-1000 cells per experiment per condition. When I run statistical tests, the difference between populations is extremely significant. As a measure of effect, I take ratio of medians of treated/untreated populations and get reasonably narrow CI. So this works well.
I repeat the experiment 2 more times, and the results are as expected - always more staining in the treated population. However, there is variation in absolute values and in the computed treated/untreated ratios. In order to demonstrate that my effect, I compute the one-sample t-test on the treated/untreated ratios and get that my treated / untreated ratio is NOT significantly different from 1. It happens because in such case N=3, in contrast to N=200 in individual experiments.
The question is, how to properly incorporate biological replicates into such analysis and not to loose the effect?
I will greatly appreciate any help!
I'm wondering what you chose as your variables to compare. Have you assigned a rank to different intensity of coloration? I think the chosen variable might be the reason for different results from statistical tests not the number of replicates. Just a thought.
Alina Avanesyan
thank you for the answer.
No, I did not assign a rank. In a single experiment I just compared medians of populations. I don't know how to incorporate the biological triplicates into the analysis.
I am new in statistics of populations, so I would really appreciate any advice.
I was thinking if you could assign a rank to cells with different coloration for each population, something like:
0 - not stained
1 - intermediate color
2 - bright color, etc.
Then, you could use any non-parametric test to see the effect of the treatment to cell's color (if this is what you want to find out). In this case, any reasonable number of replicates for each population should be fine. I hope I got your question correctly.
Thank you very much for the direction. I will look into the details. one more question - how do I treat replicates? Can I merge the data from replicates as if it was single population?
If the ultimate goal is to compare the populations I would definitely merge the data. However, if there were reasons for the replicates (e.g. certain number of cells can fit into 1 slide, 1 petri dish, etc.) then, before merging the data, it is better to compare the replicates first, to make sure there are no significant differences among them - i.e. it is better to account for variation among the slides, petri dishes, or other enclosures. Hope that will work. Best of luck!
- Badges
- Science method
- Similar topics
- Biological Science
- Histology
- Staining