Hi,

I'm using R to create graphs and run some stats. I find square root transformations are fairly easy to work with but for log transformations I'm having trouble understanding whether I'm doing it correctly or not. Is anyone able to help? Here are what my annotations look like:


> shapiro.test(final[which(Type=="DE")])
> #data not normal

> logde<-log(final[which(Type=="DE")])
> shapiro.test(logde)
> #p-value = 0.09358, transformed to normality: summarize log transformed data and back-transform to input into table

> summary(logde)
Min. 1st Qu. Median Mean 3rd Qu. Max.
-5.7860 -3.3030 -2.7090 -2.6930 -2.0630 0.3318

> mean=10^(mean(logde))
> mean
[1] 0.002029606
> median=10^(median(logde))
> median
[1] 0.001954678

> sem=sqrt(var(logde)/length(logde))
> sem
[1] 0.04730152


so far so good...right??
this is the part that I'm unsure about:

> se.m<-10^sem
> se.m
[1] 1.115068
> leftCI=mean-se.m*2
> leftCI
[1] -2.228107
> leftCI=mean(logde)-sem*2
> leftCI
[1] -2.787191
> leftCI=10^(mean(logde)-sem*2)
> leftCI
[1] 0.001632333
> rightCI=10^(mean(logde)+sem*2)
> rightCI
[1] 0.002523566
>

can someone tell me which is done the right way and why the other way is wrong??


or maybe this?
> leftCI=10^(mean(logde)-se.m*2)
> leftCI
[1] 1.194744e-05
> rightCI=10^(mean(logde)+se.m*2)
> rightCI
[1] 0.3447851
>


This post was edited by MOMOtheflyingLEMUR on Mar 6 2017 12:46pm