Skip to main content
cancel
Showing results for 
Search instead for 
Did you mean: 
Not applicable

Regression function and r squared

Hey,

do you know how to write a regression function and the r squared of the regression in the new Qlik Sense 3.2 SR3?

The way I used to do it in the old version is not working anymore.

The following is how i used to write the regression function.

R.ScriptEval('fitted.values(lm(.arg2 ~ .arg1))' , RegInnen, Avg(ADI_Skaliert))

regrads

Santana

4 Replies
sunny_talwar

This was something which worked for you in Qlik Sense?

R.ScriptEval('fitted.values(lm(.arg2 ~ .arg1))' , RegInnen, Avg(ADI_Skaliert))

Not applicable
Author

yes that worked for me.

but it is much easier with the new version of Qlik sense.

now i've got this

R.ScriptEval('lm(q$Invest ~ q$RegInnen)$fitted' , Avg(ADI_Skaliert) as Invest, RegInnen)

and it works perfectly. The only problem is that I am not able to extract the r squared of the regression for this regression

Regrads

Santana

Anonymous
Not applicable
Author

try:

R.ScriptEval('summary(lm(q$Invest ~ q$RegInnen))$r.square' , Avg(ADI_Skaliert) as Invest, RegInnen)


it brings the r square in R, but I wasn't able to handle the R return. I could  plot a bar chart with dimension equal to record number.

Anonymous
Not applicable
Author

I confirm Rodolfo answer. The result however depends where you want to have your number - in table, chart per each row of the data or in text box (recommended). In tabel or in chart you need to use rep function to repeat it in every row:

R.ScriptEval('rep(summary(lm(q$Invest ~ q$RegInnen))$r.square,length(q$RegInnen))' , Avg(ADI_Skaliert) as Invest, RegInnen). This is my result for some numbers:


In text box you shall qoute it like this (I hope it works but I do not know exact form of your data below):

Aggr(R.ScriptEval('rep(summary(lm(q$Invest ~ q$RegInnen))$r.square,length(q$RegInnen))' , Avg(Total <RegInnen> ADI_Skaliert) as Invest, RegInnen),RegInnen)

And last, why not to use qlik function LINEST_R2? In this case it might work (for fitted values R is more comfortable).