On 4 Feb 2003 at 15:27, Rado Bonk wrote:
You should really give more details, i.e. which model
did you fit? If it is a linear model with intercept, the residuals
theoretically have sum zero, and for a non-linear model the
sum should be close to zero. Comparing means of residuals
doesn't seem to make much sense.
Kjetil Halvorsen
> Hi R-users,
>
> My question is more methodological one, rather than technical.
>
> I have to samples representing residuals based on two measurements
> techniques (resid1,resid2; n=69). I need to compare two samples, to
> reject one technique (the worse one), and to keep the one which gave
> lower residuals (better one). What to look for? What should I analyse?
> Means, variance, std. deviations?
>
> Based on preliminary EDA:
>
> mean_resid1 < mean_resid2 #I would reject resid2 technique
> stdev_resid1 > stdev_resid2 #I would reject resid1 technique
> var_resid1 > var_resid2 #I would reject resid1 technique
>
> Based on means I would reject resid2 technique, but based on analyses of
> stdev. and variance I would reject resid1 technique.
>
> How to deal with that problem? What is proper statist. atrribute to look
> for? Should I use t.test()?
>
> Best regards,
>
> Rado Bonk
>
> --
> Radoslav Bonk M.S.
> Dept. of Physical Geography and Geoecology
> Faculty of Sciences, Comenius University
> Mlynska Dolina 842 15, Bratislava, SLOVAKIA
> tel: +421 905 968 127 e-mail: rbonk at host.sk
>
> ______________________________________________
> R-help at stat.math.ethz.ch mailing list
> http://www.stat.math.ethz.ch/mailman/listinfo/r-help