To validate a microarray, we used SYBR green RT PCR for 25 genes. Just wondering what the best way to confirm whether those gene expression patterns were validated.
We used 4 different treatments and 3 different donors.
Looking at them, many genes look almost identical. I would prefer statistics to either justify or refute this.
If you are interested in a similarity of patterns, then just do a correlation. You can also use a Bland-Altman plot to show the consistency between the two assays. A hypothesis test requires a point-null hypothesis. A thinkable null hypothesis was the identity of values between the assays, what clearly is non-sensical. Further, a test may make you reject the null, what means that you don't believe in the equality of the patterns. Failing to reject the null, on the other hand, tells you nothing (i.e., you may still believe both, equality and inequality of the patterns; the "non-significant" test result just does not help you to find a decision). I would argue with the *picture* (scatterplot or Bland-Altman plot), not with some arbitrary test statistic.
If you really really want some p-value, I would test the null the *directions* of regulation are uncorrelated. In this case, the proportion of equal signs in regulation should be 0.5. The test is a simple binomial test.
- Badges
- Science topic