linear regression: strange values

zozette

Hi all,
I have a database. I have conducted a linear regression with Gretl, R, Orange data mining and RapidMiner.
I get the same results with Gretl, R, Orange data mining but RapidMiner gives me very strange results.
Look at std error, p-value and t-value.
https://www.dropbox.com/sh/ab2tt2uq1klfbux/tUnIrDTbfZ
Is this a bug ?

Milane

aborg

I have similar problem. Help is appreciated.

MariusHelf

Hi Milane,

can you describe what is strange about the results? If I compare the RapidMiner results it to the results above it, I see no significant differences.

Best regards,
Marius

zozette

Marius wrote:
Hi Milane, can you describe what is strange about the results? If I compare the RapidMiner results it to the results above it, I see no significant differences. Best regards, Marius

Look at standard error, t-value or p-value, for example, for the intercept.
It seems that the correct value for std error is 2.73 and no 2.222
for the t-value 0.356 and no 0.438
for the p-value 0.723 and no 0.667
These are significant differences 
In RapidMiner, are these values calculated in the usual way? If so, the results should be the same, no?