RSS Residual Sum Squares

#1
Hi

im analysing my dissertation which has a OLS model. My RSS figures are very very big in some cases above 15000. I know a high RSS indicates a poorly fitted model. i have run the regression for 6 different countries using 154 observations per variable using 3 variables.

My dependant variable is CPI. my explanatory variables are gold prices, crude oil prices and food prices.

please help!!
 

hlsmith

Omega Contributor
#2
I don't use linear reg much, but it is the residual sum of squares, so if you are working with big numbers it will be big. Given that, 15,000 may not be an indicator of poor fit, since I believe it is contextual.


Thanks.
 

noetsi

Fortran must die
#3
The best way to decide if it is big is to try to find in the literature what other models have. Of course the problem with that is you may not find much in the literature or they may not report this value.
Another way to determine if its too high is too look at other measures of fit in linear regression.