Hi everyone, I have some trouble finding the correct method for statistical analysis. I was thinking about a two-tailed paired T test, but that only considers the mean value of my replicates and not the distribution of the individual replicates as well.
My data set consists of 4 groups that are divided based on percentages (together 100%).
These groups are dependent on one variable (control, A, B, C, D, E and F) and I want to know whether condition A, B, C etc. is significantly different from the control.
I have 3 replicates of the experiment (with some measurement variance).
Since you have a control variable then you can use ANCOVA and control for each and every one of them over ther to see if the actually covary in the real sense. Mind you, you are using the control group to check.
If t-tests make sense depends on the kind of response variable. If the normal distribution is a resonable model for the distributions of the response values, thaen t-tests are ok. Then, Dunnett's procedure is a solution tailored for your problem, controlling the family-wise error-rate in a series of "multiple-to-one" comparisons by t-tests using a pooled estimate of the standard error.
https://en.wikipedia.org/wiki/Dunnett%27s_test
- Badges
- Science method