[BioC] mt.teststat (multtest)

[Ann Hess]

I apologize for the foolish mistake.  Of course, the usual t-statistic 
(assuming equal variances) is equal
to Welch's t-statistic (assuming differing variances) if n_1 = n_2, i.e.
when both groups have the same size.

Ann

---------- Forwarded message ----------
Date: Tue, 17 Jan 2006 18:15:14 +0100 (MET)
From: Holger Schwender <holger.schw at gmx.de>
To: Ann Hess <hess at stat.colostate.edu>
Subject: Re: [BioC] mt.teststat (multtest) BUG?

Hi Ann,

that's not a bug. The usual t-statistic (assuming equal variances) is equal
to Welch's t-statistic (assuming differing variances) if n_1 = n_2, i.e.
when both groups have the same size.

Best,
Holger


> --- Ursprüngliche Nachricht ---
> Von: Ann Hess <hess at stat.colostate.edu>
> An: BioConductor Mail List <bioconductor at stat.math.ethz.ch>
> Betreff: [BioC] mt.teststat (multtest) BUG?
> Datum: Tue, 17 Jan 2006 09:20:24 -0700 (MST)
>
> I am using the function mt.teststat (from multtest package) in R2.2 with
> BioC 1.7.
>
> When I use mt.teststat(test="t") or mt.teststat(test="t.equalvar") I get
> the
> same results (which assumes EQUAL variances).  See code below.
>
> Am I making some mistake or could there be a bug in the function?
>
> Ann
>
> **************************
> cl<-c(1,1,0,0,1,0)
> test1<-mt.teststat(exprs(data),cl,test="t")
> test2<-mt.teststat(exprs(data),cl,test="t.equalvar")
>
> test1[1:10]
> # -1.7941959 -0.8467331 -1.0077421 -0.3012158  2.4909231
> test2[1:5]
> # -1.7941959 -0.8467331 -1.0077421 -0.3012158  2.4909231
>
> #Variances are NOT EQUAL:
> var1.1<-var(exprs(data)[1,c(1:2,5)])
> var1.2<-var(exprs(data)[1,c(3:4,6)])
> var1.1
> # 43
> var1.2
> # 305.33
>
> #Test statistic uses POOLED t-test:
> mean1.1<-mean(exprs(data)[1,c(1:2,5)])
> mean1.2<-mean(exprs(data)[1,c(3:4,6)])
> sp.1<-sqrt( (2*var1.1 + 2*var1.2)/4 )
> t.1<- (mean1.1-mean1.2)/(sp.1*sqrt(2/3))
> t.1
> # -1.794196
>
> _______________________________________________
> Bioconductor mailing list
> Bioconductor at stat.math.ethz.ch
> https://stat.ethz.ch/mailman/listinfo/bioconductor
>

-- 
10 GB Mailbox, 100 FreeSMS/Monat http://www.gmx.net/de/go/topmail
+++ GMX - die erste Adresse für Mail, Message, More +++