Posted 17 February 2012 - 05:38 AM
Posted 14 November 2012 - 05:33 AM
This is primarily done on the dCt values, but we also do it for the 2-ddCt values for comparison,
It is in our animal data generally so that the dCT values are more often normal distributed as compared to the 2-ddCt values,
At present (and I am not sure we will continue doing this because I am also in the process of getting a better understanding of this) we then do the ANOVA (or the non-parametric Kruskal Wallis test in case the data are not normal distributed) on the dCt values. Graphically we still show the graphs as 2-ddCt. If I at a certain point would choose to do the statistics on the 2-ddCT values because of better normal distribution, I would probably actually do it on the 2-dCt values to have a control group in there for the ANOVA.
I am not super enthusiastic about the delta delta method, but I can see some advantages as compared to the standard curve method (which I am used to from my former labs).
Posted 15 November 2012 - 04:46 AM
Actually yesterday I re-read a Comparative Ct method paper by Schmittgen and Livak (Nature protocols, vol 3. no. 6 2008 page 1101), and they recommend to do the statistics on the 2-Ct (for testing of whether the housekeeping gene is appropriate to use), or the 2-dCt (for data where there is no link between the control and treated sample) or the 2-ddCt values (where there is a link between the control and the treated sample; e.g. if the control is a period without treatment in a human and the treated sample is form a treatment in the same person).
And that prompted me to look at several datasets that we are currently evaluating. For these (also when considering tendency for the dCt data to be normal distributed), I decided to use the 2-dCt values for statistics, and in case of normality to run a one-way ANOVA. I also decided for these data to represent the data graphically as the dCt values as percent of control.