An embedded and charset-unspecified text was scrubbed... Name: not available Url: https://stat.ethz.ch/pipermail/bioconductor/attachments/20051108/ 465c7096/attachment.pl

As another poster have said: you are looking at interactions. Basically you have a scale problem here: you have a synergistic effect on one scale but not on the other. Because log converts multiplication to addition, a synergistic effect on a log scale is different than a synergistic effect on the original scale. From a naive point of view, this is bewildering. You need to be precise wrt. what you mean by a synergistic effect and then realize that this definition is not scale independent (however strange that may seem). Kasper On Nov 8, 2005, at 8:55 AM, Lance Palmer wrote: > I am interested to know how one may determine if two conditions can > have > synergistic effects. For example, lets say we expose cells to two > conditions, X and Y. We have four microarray chips. > > > > 0 X Y X+Y > > > > Lets suppose the following raw data > > > > 8 2046 16 4096 > > > > Log2 of those numbers > > > > 3 11 4 12 > > > > Log fold change vs no stimuli > > > > 8 1 9 > > > > Fold change > > > > 256 2 512 > > > > Raw change > > > > 2038 8 4088 > > > > > > > > If you look at the fold change one can say there is no synergistic > effect > since the log fold change of 9 can be explained by the 2**8 > increase and the > 2**1 increase. (or I could say the 512 comes from the 256x > component and 2x > component) > > > > If you look at the raw change, it appears to be synergistic since > 2038+8<<<4088. > > > > I am not a mathematician so I don't know which would be the proper > way to > look at this. Would an example like this be synergistic. And what > is the > best way to approach this. > > > > > > > [[alternative HTML version deleted]] > > _______________________________________________ > Bioconductor mailing list > Bioconductor at stat.math.ethz.ch > https://stat.ethz.ch/mailman/listinfo/bioconductor

What you are looking for is called "interaction" in statistics and the design of your experiment is described as a 2x2 factorial design. This is also sometimes described as two-way ANOVA (analysis of variance). You will need to ensure that the data follows a normal distribution (e.g. by taking log2) in order to use this method as it assumes normally distributed error terms. Trying reading up http://www.uwsp.edu/psych/stat/13/anova-2w.htm http://www.bioconductor.org/repository/devel/vignette/factDesign.pdf PS : Another reason why statisticians like logged values is that it is symmetric around 1. E.g. raw values of 0.5 and 2.0 are -0.693 and 0.693 on log scales. Regards, Adai On Tue, 2005-11-08 at 11:55 -0500, Lance Palmer wrote: > I am interested to know how one may determine if two conditions can have > synergistic effects. For example, lets say we expose cells to two > conditions, X and Y. We have four microarray chips. > > > > 0 X Y X+Y > > > > Lets suppose the following raw data > > > > 8 2046 16 4096 > > > > Log2 of those numbers > > > > 3 11 4 12 > > > > Log fold change vs no stimuli > > > > 8 1 9 > > > > Fold change > > > > 256 2 512 > > > > Raw change > > > > 2038 8 4088 > > > > > > > > If you look at the fold change one can say there is no synergistic effect > since the log fold change of 9 can be explained by the 2**8 increase and the > 2**1 increase. (or I could say the 512 comes from the 256x component and 2x > component) > > > > If you look at the raw change, it appears to be synergistic since > 2038+8<<<4088. > > > > I am not a mathematician so I don't know which would be the proper way to > look at this. Would an example like this be synergistic. And what is the > best way to approach this. > > > > > > > [[alternative HTML version deleted]] > > _______________________________________________ > Bioconductor mailing list > Bioconductor at stat.math.ethz.ch > https://stat.ethz.ch/mailman/listinfo/bioconductor >