hi,
to help assess the validity of the data and output from linear regression modelling etc, it would be useful to determine how the residuals change relative to the independent variable in a residual graph.
this can help determine whether a non linear regression would be better, or even whether data needs transforming into logs etc.
link below shows typical examples and usage.
http://stattrek.com/m/regression/residual-analysis.aspx?tutorial=ap
simon.
Hi ,
I totally agree with you. :)
I am new to KNIME, but after spending a lot of time looking around in KNIME, I relaized that there is no function for residual analysis of regression errors. The regression icon does calculate the errors and places them in a column at the end of the data file, but for some weird reason, they calculate the absolute value.... So, the only way, as far as I can tell, is to eye-ball the fitted curve, or plot the estimated depended against the actual one, and eye-ball them.