Dear R-devel,
It seems that lsfit incorrectly reports coefficients when the input matrix 'x'
is rank-deficient, see the example below:
## here values of 'b' and 'c' are incorrectly swapped
> x <- cbind(a=rnorm(100), b=0, c=rnorm(100)); y <- rnorm(100); lsfit(x, y)$coef
Intercept a b c
-0.0227787 0.1042860 -0.1729261 0.0000000
Warning message:
In lsfit(x, y) : 'X' matrix was collinear
## correct values
> lsfit(x[,-2], y)$coef
Intercept a c
-0.0227787 0.1042860 -0.1729261
I looked inside the lsfit code and it appears that even though rank-deficiency
is detected there is no attempt to patch the coefficients. Why is that?
Taking clues from the code it appears that the following trick might do the
work:
> foo <- lsfit(x, y)
Warning message:
In lsfit(x, y) : 'X' matrix was collinear
> structure(foo$coefficients[foo$qr$pivot], names=names(foo$coefficients))
Intercept a b c
0.14857345 -0.07473099 0.00000000 0.12835155
Is this reliable or there are cases when it may fail?
Thanks,
Vadim
P.S.
> version
_
platform i386-pc-mingw32
arch i386
os mingw32
system i386, mingw32
status
major 2
minor 7.1
year 2008
month 06
day 23
svn rev 45970
language R
version.string R version 2.7.1 (2008-06-23)
>
Note: This email is for the confidential use of the named addressee(s) only and
may contain proprietary, confidential or privileged information. If you are not
the intended recipient, you are hereby notified that any review, dissemination
or copying of this email is strictly prohibited, and to please notify the
sender immediately and destroy this email and any attachments. Email
transmission cannot be guaranteed to be secure or error-free. Jump Trading,
therefore, does not make any guarantees as to the completeness or accuracy of
this email or any attachments. This email is for informational purposes only
and does not constitute a recommendation, offer, request or solicitation of any
kind to buy, sell, subscribe, redeem or perform any type of transaction of a
financial product.
______________________________________________
R-devel@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-devel
