1

I am doing a gene expression analysis on blood samples collected from the same 8 goats. My sampling points are before LPS-treatment (0h), and 1h, 2h, 5h and 24h after treatment. I want to compare the mean expression of each post-treatment to the before-treatment values. Due to problems in RNA-extraction, one sample is missing at 1h and one sample is missing at 2h. Is it possible to do a paired t-test with unequal sample size? E.g n=8 at 0h and n=7 at 1h/2h?

I'm using log2 transformed data for the statistical analysis. (We are have decided not to correct for multiple comparisons, but we use a fold-change limit instead)

kjetil b halvorsen
  • 63,378
  • 26
  • 142
  • 467
Øyvind S
  • 11
  • 1
  • Asked&answered multiple times before, specifically https://stats.stackexchange.com/questions/74960/missing-data-spss-paired-samples-t-test, https://stats.stackexchange.com/questions/379085/paired-t-test-following-main-effect-of-mixed-model-with-missing-data, https://stats.stackexchange.com/questions/217535/paired-t-test-for-repeated-measurements-but-with-2-missing-samples and https://stats.stackexchange.com/questions/147309/missing-data-paired-samples-t-test, https://stats.stackexchange.com/questions/282525/paired-repeated-measures-anova-with-missing-cases-or-mixed-model – kjetil b halvorsen Jul 08 '19 at 08:39

1 Answers1

0

You can not do a t-test for paired data with unequal sample-sizes but you can omit cases with missing data for the analysis of these times: Do a n=8 comparison for 5h an 24 h and an n=7 comparison for n=1h an n=2h.

Have you considered repeated-measurements-ANOVA instead of all those t-tests? With n so small and 4 different tests, alpha Inflation is an issue to be considered.

Bernhard
  • 7,419
  • 14
  • 36