Ddear colleagues
I'm performing a statistical analysis of real-time-PCR data by the delta delta ct method. Basically I'm analysing the expression profile of a target gene in non-inoculated (control) and infected samples, at several time points. My approach would be multifactorial anova followed by Fisher's LSD. However, I think that LSD can only be performed if population variances are equal (homogeneity of variance ). I'm using SPSS software and when I do a anova>LSD it also gives me the levene's test for (homogeneity of variance ) (levene's>anova>LSD). In my data the p value is less than 0.05 and the null hypothesis is rejected and therefore, the homogeneity of variance is rejected, and therefore, the LSD can not be applied. Is this so?
I have been reading some papers and in the statistical analysis section, is always referred the anova and LSD but no Levene's test or homogeneity of variance test.
If I know that levene's test fails for homogeneity of variance of my sample, what tests can I apply? Thanks
What is the data you are analyzing?
PS: using formal tests to justify the "reasonability of assumptions" (like Levène's test for the variance homogeneity or Shapiro-Wilk for the normal distribution) is logically absurd. If such tests lead you to reject the tested H0, you don't know if the statistical significance of the data under H0 is of any relevance, and if you don't reject, you don't know if the power was too small to detect relevant deviations from H0.
To my understanding you can do ANOVA and Tukey test, but selection of reference gene requires homogeneity of variance to be tested by leavenes test, where in the
In continuation to the above answer, the homogeneity of variance should be tested for the reference gene and then your target gene should be normalized with the selected reference gene.
- Badges
- Science topic