One-way ANOVA

Set the variable nt equal to the number of treatments.

Put your data in arrays data[0], data[1], etc. and hit the Go! button.

Input

nt=3
data = new Array(nt)

data[0] = [102,101,101,100,102]
data[1] = [99,100,99,101,98]
data[2] = [103,100,99,104,102]

n=new Array(); mean=new Array(); sd=new Array()
res=3

tsumx=0; tsumx2=0; N=0; wiss=0
for(i=0;i<nt;i++) {
n[i]=data[i].length
N+=n[i]
sumx=0; sumx2=0;
   for(j=0;j<n[i];j++) {
   sumx+=data[i][j]; tsumx+=data[i][j]
   sumx2+=data[i][j]*data[i][j]; tsumx2+=data[i][j]*data[i][j]
   }
mean[i]=sumx/n[i]
variance=sumx2-sumx*sumx/n[i]; wiss+=variance
sd[i]=sqrt(variance/(n[i]-1))
output+="Mean of Group " + (i+1) + ": " + mean[i].toFixed(res)+"\n"
// output+="Sum of Squares of Group " +i+ ": " + variance+"\n"
   }
tmean=tsumx/N
output+="Overall Mean: "+ tmean.toFixed(res) +"\n"
tdf=N-1
output+="Total Degrees of Freedom: "+tdf+"\n"
tss=tsumx2 - tsumx*tsumx/N
output+="Total Sum of Squares: "+tss.toFixed(res)+"\n"
btdf=nt-1
output+="Between Group Degrees of Freedom: "+btdf+"\n"
btss=tss-wiss
output+="Between Group Sum of Squares: "+btss.toFixed(res)+"\n"
btms=btss/btdf
output+="Between Group Mean Square: "+btms.toFixed(res)+"\n"
widf=tdf-btdf
output+="Within Group Degrees of Freedom: "+widf+"\n"
output+="Within Group Sum of Squares: "+wiss.toFixed(res)+"\n"
wims=wiss/widf
output+="Within Group Mean Square: "+wims.toFixed(res)+"\n"
F=btms/wims
output+="F: "+F.toFixed(res)+"\n"
m=btdf; n=widf
fcrit=74.2091305*pow(n,-2.115079986)+2.639078663*pow(m,-0.670693668)+1.627481284
output+="Approx. F-crit (p=0.05): "+fcrit.toFixed(1)+"\n"
if(F>fcrit) output+="A significant difference exists\n"
else output+="No significant difference exists\n"

Output

You can Save your data in a cookie and Load it later on the same computer.

The calculation uses a very crude estimate of the F-distribution, so don't base you life on the results.

Email. July 4, 2009