Re: [R] OLS error

From: Roger Bivand <>
Date: Fri 12 Nov 2004 - 19:46:15 EST

On Thu, 11 Nov 2004, DrakeGis wrote:

> Thanks for the help.
> -I'm really sorry, but I'm affraid I can't publish any data in order to
> allow a reproduction of the results (enterprise policies :( ).

The author of the Design package has already replied to this point. If the function only fails with your data, the data are the problem, not the function.

> -I'm using ols() from Design (Is there another ?)

The posting guide does say read the appropriate documentation. R is distributed with lots of documentation, including "An Introduction to R", see -> Statistical models in R -> Linear models. Reading the help page for the ols() function from the Design package will show that it adds functionality to the base lm() function, which your example is not using, so you can certainly use lm() until you need more.

> -Does version of R influence on the behavior of the packages ? I didn't
> know, I'm using Version 2.0.0 (2004-10-04), ISBN 3-900051-07-0

Yes, but this isn't the case here. If you had read the posting guide, you would understand that some questions are easier to answer when the version is known, and the version is found by typing version at the R prompt, surprisingly, and shows basic information about the OS too.

> > So in addition to telling us which version of R you are using, you could
> > say which package you are using, and provide a reproducible example.
> > If you look at help(La.chol2inv), you will see that the function takes
> > arguments: La.chol2inv(x, size = ncol(x)), where the x provided within the
> > ols() function with your data has only one row. Could you perhaps make a
> > data frame out of your variables, and use the data= argument? Are there
> > many missing values in your variables? Have you tried complete.cases() on
> > your input data?
> -I have no missing values. I previously try to use a data.frame but the
> was other error message, and because of that I split the data.frame,
> finally after you suggestions, I create a new data.frame and still didn't
> work until I attached a the new data.frame and it works.

Please copy the contents of the console to your messages. Any of this could be caused by simple mistakes, attach() is not needed often (I never use it), and especially not for lm() and friends, which have the data= argument to be sure that the object in the formula is the one in the data= object, not any other with the same name that might be lying around. Try running in a fresh session too without loading your old workspace - one or other of your RHS variables is not what you think it is.

There is very little doubt that you have problems with your data. However, you have not bothered to show the error messages to the list. Try to go through the same steps using lm() and making the data from runif() and rnorm() - simulated data using set.seed() - and then run the examples for lm(), etc, and if you can demonstrate that the functions still return errors, the cause should be easier to establish.

> Thanks
> D.
> -----------------------------------------
> Stay ahead of the information curve.
> Receive GIS news and jobs on your desktop daily.
> Subscribe today to the GIS CafeNews newsletter.
> [ ]
> It's informative and essential.

Roger Bivand
Economic Geography Section, Department of Economics, Norwegian School of
Economics and Business Administration, Breiviksveien 40, N-5045 Bergen,
Norway. voice: +47 55 95 93 55; fax +47 55 95 93 93

______________________________________________ mailing list
PLEASE do read the posting guide!
Received on Fri Nov 12 19:59:26 2004

This archive was generated by hypermail 2.1.8 : Fri 12 Nov 2004 - 21:08:15 EST