[R] inconsistent behaviour of add1 and drop1 with a weighted linear model

[Duncan Murdoch]

On 3/13/2007 7:46 PM, hadley wickham wrote:
> On 3/13/07, Duncan Murdoch <murdoch at stats.uwo.ca> wrote:
>> On 3/13/2007 3:14 PM, hadley wickham wrote:
>>> On 3/13/07, Peter Dalgaard <p.dalgaard at biostat.ku.dk> wrote:
>>>> Jenny Hodgson wrote:
>>>>> I was using Version 2.3.1 for Windows (binary download version). Guess I
>>>>> should be using the latest version, (but the reason is I'm writing up my
>>>>> PhD and I thought my results would be more 'repeatable' if I didn't keep
>>>>> changing my version of the software, I didn't really think there would
>>>>> be any glitches as big as this). Sorry if this is a waste of your time.
>>>>> And thanks very much for replying so quickly.
>>>>>
>>>>> Jenny
>>>>>
>>>>>
>>>> Always a good idea to check the NEWS file:
>>>>
>>>>     o   add1.lm() had been broken by other changes for weighted fits.
>>> Yes, but a little impractical.  Has any one ever considered a standard
>>> for the news file so changes could be extracted and included in online
>>> documentation somewhere?
>> It's in a structured format, and is online.  The readNEWS() function can
>> read it.  It would be nice if someone would contribute a more friendly
>> reader...
>>
>> Try
>>
>> readNEWS(url("http://cran.r-project.org/bin/windows/base/NEWS.R-2.5.0dev"),
>> chop="keepAll")[[c("2.5","2.5.0","BUG FIXES")]]
>>
>> for the unfriendly but informative version.
> 
> Oh, that's neat.  One thing that would be very useful would be to
> figure out what (if any) functions an news item refers to.  A grep
> against all function names in base packages wouldn't be too hard to
> do.

I think we are reasonably consistent in referring to functions with 
parens following, e.g. add1.lm() above, so a search for patterns like 
that should find most of them.

Duncan Murdoch

> 
> I'll think about it, and how to render it nicely in a webpage.
> 
> Hadley