The F ratio compares the variability between groups to a supposedly
independent estimate of what that variability would be if there were no
group effect.
If F is large (much greater than 1), it suggests that there probably
is a group effect.
If F is small (quite close to 0), it suggests some other violation of
assumptions. This can occur in a balanced 2-way layout where factor A
is unimportant while B is large but excluded from the model.
tst.df2 <- data.frame(A=rep(letters[1:2], each=2), B=rep(letters[1:2],
2), y = rep(1:2, 2)+0.01*rnorm(4))
anova(lm(y~A, tst.df2))
Analysis of Variance Table
Response: y
Df Sum Sq Mean Sq F value Pr(>F)
A 1 0.00010 0.00010 2e-04 0.9896
Residuals 2 0.96970 0.48485
Hope this helps
Spencer Graves
Rishabh Gupta wrote:> Hello,
> I have a question regarding anova. I am a bit comfused about the F value.
My understanding is
> that the F value indicates the ability of numerical variable V to
discriminate between different
> values of grouping variable G. If the F value is equal or close to 1.0, it
means that all the
> variances of the groups are similar and that V is unable to discriminate
for G. On the other hand
> if the F value is far from 1.0, it suggests that the variances of the
groups are different.
> The issues that I am comfused about is F values of <1 vs >1. My
understanding is that the values
> are equal to their reciprocal as far as the discriminance is concerned. So
0.5 is equivalent to
> 1/0.5. Could somebody please tell me whether this is correct or not, or
whether I have got it
> completely wrong.
> Your help is greatly apprecited. Many Thanks
>
> Rishabh
>
> ______________________________________________
> R-help at stat.math.ethz.ch mailing list
> https://www.stat.math.ethz.ch/mailman/listinfo/r-help