Diff for "FAQ/power/props" - CBU statistics Wiki
location: Diff for "FAQ/power/props"
Differences between revisions 2 and 3
Revision 2 as of 2006-09-14 15:32:08
Size: 1540
Editor: PeterWatson
Comment:
Revision 3 as of 2006-09-14 15:37:16
Size: 1909
Editor: PeterWatson
Comment:
Deletions are marked like this. Additions are marked like this.
Line 1: Line 1:
[ADJUST THE EXAMPLE INPUT AS DESIRED; THE COPY AND PASTE INTO A SPSS SYNTAX WINDOW AND RUN; OUTPUT BOTH TO SPREADSHEET AND OUTPUT WINDOW].
Line 2: Line 3:
The effect size used here is the Odds Ratio See
http://en.wikipedia.org/wiki/Effect_size#Cohen.27s_d which can also be expressed as a difference in proportions.
This program uses the difference in proportions (or equivalently the odds ratio) as the effect size, see http://en.wikipedia.org/wiki/Effect_size#Cohen.27s_d.
Line 5: Line 5:
Alpha is the (two-tailed) type I error, p1 and p2 are the proportions in groups 1 and 2being compared, ratio is the ratio of sizes of group 1 to group 2 and ntot is the total sample size. The program then outputs the power.

 

[ADJUST THE EXAMPLE INPUT AS DESIRED; THE COPY AND PASTE INTO A SPSS SYNTAX WINDOW AND RUN; OUTPUT BOTH TO SPREADSHEET AND OUTPUT WINDOW].

This program uses the difference in proportions (or equivalently the odds ratio) as the effect size, see http://en.wikipedia.org/wiki/Effect_size#Cohen.27s_d.

Alpha is the (two-tailed) type I error, p1 and p2 are the proportions in groups 1 and 2being compared, ratio is the ratio of sizes of group 1 to group 2 and ntot is the total sample size. The program then outputs the power.

DATA LIST free
/p1 p2 ratio ntot alpha.
BEGIN DATA.
0.19 0.09 1 311 0.05
0.19 0.09 4 311 0.05
0.19 0.09 3 600 0.05
END DATA.

compute p11= (p1*ratio)/(1+ratio).
compute p12=((1-p1)*ratio)/(1+ratio).
compute p21=p2/(1+ratio).
compute p22=(1-p2)/(1+ratio).

compute oddsr = (p11*p22)/(p21*p12).
exe.

matrix.
get m /variables=p11 p12 p21 p22 ntot alpha /missing=omit.
compute p11=make(1,1,0).
compute p12=make(1,1,0).
compute p21=make(1,1,0).
compute p22=make(1,1,0).
compute alpha=make(1,1,0).
compute p11=m(:,1).
compute p12=m(:,2).
compute p21=m(:,3).
compute p22=m(:,4).  
compute ntot=m(:,5).
compute alpha=m(:,6).
end matrix.
COMPUTE #chisq1 = (LN(oddsr)/(sqrt((1/p11)+(1/p12)+(1/p21)+(1/p22))))**2.
compute #df=1.
compute #conf = 1-alpha.
compute #chisq=ntot*#chisq1.
compute power = 1 - NCDF.CHISQ(IDF.CHISQ(#conf,#DF),#DF,#CHISQ).
compute p1=p11/(p11+p12).
compute p2=p21/(p21+p22).
compute ratio = (p11+p12)/(p21+p22).
formats p1 (f5.2) p2 (f5.2) alpha (f5.2) ntot (f5.2) power (f5.2).
variable labels p1 'Proportion 1' /p2 'Proportion 2' /ratio 'Ratio group 1 to group 2' /alpha 'Two-Tailed Alpha' /ntot 'Sample Size' /power 'Power'.
report format=list automatic align(center)
  /variables=p1 p2 ratio alpha ntot power 
  /title "Power for given sample size for an Odds Ratio from a 2x2 table of proportions" .

None: FAQ/power/props (last edited 2013-03-08 10:17:40 by localhost)