Discussion:
Problem with linear regression in PSPP 0.10.5-pre2
(too old to reply)
Dr. Oliver Walter
2017-08-27 09:04:09 UTC
Permalink
Dear all,

I came across a problem with the parameter estimates variance-covariance
matrix of linear regression in PSPP 0.10.5-pre2g9a68ff on a Windows 10
computer. This matrix is not correctly printed in the output, as can be
seen on the attached photos (pspp vs. results from R): 2.480, for
example, is not the error variance of the variable "horsepower", but of
the variable "type", as can be clearly seen if you compare it with the
standard errors of the parameter estimates above. 0.0135 seems to be the
covariance between "horsepower" and "type" and not between "horsepower"
and "engine size".

Kind regards,

Oliver Walter
John Darrington
2017-08-28 10:52:45 UTC
Permalink
I will investigate this asap, but I am catching up on a lot of other work.

So that it doesn't get forgotten, could you raise is as a bug at
https://savannah.gnu.org/bugs/?group=pspp

and please include a copy of the syntax and dataset which provokes the problem.

Thank you.

On Sun, Aug 27, 2017 at 11:04:09AM +0200, Dr. Oliver Walter wrote:
Dear all,

I came across a problem with the parameter estimates variance-covariance
matrix of linear regression in PSPP 0.10.5-pre2g9a68ff on a Windows 10
computer. This matrix is not correctly printed in the output, as can be
seen on the attached photos (pspp vs. results from R): 2.480, for
example, is not the error variance of the variable "horsepower", but of
the variable "type", as can be clearly seen if you compare it with the
standard errors of the parameter estimates above. 0.0135 seems to be the
covariance between "horsepower" and "type" and not between "horsepower"
and "engine size".

Kind regards,

Oliver Walter
--
Avoid eavesdropping. Send strong encrypted email.
PGP Public key ID: 1024D/2DE827B3
fingerprint = 8797 A26D 0854 2EAB 0285 A290 8A67 719C 2DE8 27B3
See http://sks-keyservers.net or any PGP keyserver for public key.
Continue reading on narkive:
Loading...