[Rd] weights in glm (PR#8720)

ripley at stats.ox.ac.uk ripley at stats.ox.ac.uk
Tue Mar 28 19:39:32 CEST 2006


First, R-bugs is not for asking questions, only for reporting things you 
are *certain* are bugs: see the R FAQ.

Why are you using weights to omit cases?  If you had used subset, this 
would have worked.  The problem is the use of zero weights, which are not 
intended to be used in this way.  We can fix this up, but it is not the 
correct way to use glm.


On Tue, 28 Mar 2006, robert.pusz at wp.pl wrote:

> Full_Name: Robert Pusz
> Version: 2.2.1
> OS: Windows
> Submission from: (NULL) (157.25.9.126)
>
>
> Hello,
> In my opinion something is wrong with 'weights' option in glm.
> My code is following:
> ###begin of the code####
> cl <- c(5012, 106, 3410, 5655, 1092, 1513, 557, 1351, 3133, 2063, 3257,
> 4179, 5582, 5900, 8473, 4932, 3463, 5596, 2262, 0, 2638, 1111,
> 4881, 4211, 6271, 5257, 6926, 6165, 0, 0, 898, 5270, 2268, 5500,
> 6333, 1233, 1368, 0, 0, 0, 1734, 3116, 2594, 2159, 3786, 2917,
> 0, 0, 0, 0, 2642, 1817, 3479, 2658, 225, 0, 0, 0, 0, 0, 1828,
> 0, 649, 984, 0, 0, 0, 0, 0, 0, 599, 673, 603, 0, 0, 0, 0, 0,
> 0, 0, 54, 535, 0, 0, 0, 0, 0, 0, 0, 0, 172, 0, 0, 0, 0, 0, 0,
> 0, 0, 0)
>
> w <- c(1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0,
> 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 1,
> 1, 1, 1, 1, 1, 0, 0, 0, 0, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 1, 0,
> 1, 1, 0, 0, 0, 0, 0, 0, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 1, 1, 0,
> 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0)
>
> row <- c(1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10,
> 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10,
> 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10,
> 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10,
> 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10)
>
> col <- c(1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
> 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 5,
> 5, 5, 5, 5, 5, 5, 5, 5, 5, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 7, 7,
> 7, 7, 7, 7, 7, 7, 7, 7, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 9, 9, 9,
> 9, 9, 9, 9, 9, 9, 9, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10)
> Row <- ordered(as.factor(row))
> Col <- ordered(as.factor(col))
> fit <- glm(cl ~ Row + Col, family = quasipoisson, weights = w)
> ###end of the code####
>
> When I have written summary(fit) I have got:
>            Estimate Std. Error t value Pr(>|t|)
> (Intercept)  7.64459         NA      NA       NA
> Row2        -0.05467         NA      NA       NA
> Row3         0.24541         NA      NA       NA
> Row4         0.42035         NA      NA       NA
> Row5         0.43961         NA      NA       NA
> Row6         0.04532         NA      NA       NA
> Row7        -0.04881         NA      NA       NA
> Row8         0.25370         NA      NA       NA
> Row9        -0.14976         NA      NA       NA
> Row10       -0.01267         NA      NA       NA
> Col2         0.69283         NA      NA       NA
> Col3         0.62603         NA      NA       NA
> Col4         0.27695         NA      NA       NA
> Col5         0.06056         NA      NA       NA
> Col6        -0.19582         NA      NA       NA
> Col7        -0.83044         NA      NA       NA
> Col8        -1.27914         NA      NA       NA
> Col9        -1.93235         NA      NA       NA
> Col10       -2.49709         NA      NA       NA
>
> When I omited 'weights=w' above table was filled in with numbers, but the
> results were wrong (because of taking zeros in regression).
>
> Could you tell me what's wrong?
> Kind regards,
> Robert
>
> ______________________________________________
> R-devel at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-devel
>
>

-- 
Brian D. Ripley,                  ripley at stats.ox.ac.uk
Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
University of Oxford,             Tel:  +44 1865 272861 (self)
1 South Parks Road,                     +44 1865 272866 (PA)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595



More information about the R-devel mailing list