[R] HMisc describe -- error with dates

[tmurph6@po-box.mcgill.ca]

Thank you Gabor and Frank for you replies last week. Unfortunately I still don't
get it (I'm kind of new at this). Just to recap: I'm starting with a Julian
date and want results presented as 11NOV2003, for example (or some other date
format with no time component). 

The HMisc describe function accepts variables that I have now formatted like
this   

>date.ddmmmyy(aidsdate)

whereas it does not like variables formatted like this

>dates(aidsdate,format="day.mon.year", origin=c(month = 1, day = 1, year =
1960))

which is what I was using with an earlier version of Hmisc.

I wasn't able to figure out how to use posix.

The problem now is that, unlike before, the dates are not sorted (at least not
chronologically--maybe like character variables) and I liked seeing the 5
lowest and highest dates in the describe output. 

Sorry for pestering, but it is a testimony to how much I like Hmisc :)

Thanks!

Tanya



Quoting Frank E Harrell Jr <feh3k at spamcop.net>:

> On Tue, 11 Nov 2003 20:40:50 -0500 (EST)
> "Gabor Grothendieck" <ggrothendieck at myway.com> wrote:
> 
> > 
> > 
> > dates is part of chron.  There may be something wrong in 
> > Hmisc's detection of chron objects.
> > 
> > > chron(1)
> > [1] 01/02/70
> > 
> > > isChron(chron(1))
> > FALSE
> > 
> > ---
> >  
> > Date: Tue, 11 Nov 2003 20:06:38 -0500 
> > From: <tmurph6 at po-box.mcgill.ca>
> > To: Frank E Harrell Jr <feh3k at spamcop.net> 
> > Cc: <r-help at stat.math.ethz.ch> 
> > Subject: Re: [R] HMisc describe -- error with dates 
> > 
> >  
> >  
> > I am using the chron package. I have no preference for which function I
> > use. I just want the most reliable format for HMisc functions and
> > general plotting. I like the ddmonyyy formats (e.g. 11NOV2003). What
> > would you recommend?
> 
> I like POSIX but chron should be workable.  The fix for isChron in Hmisc
> is:
> 
> function(x) {
>   cl <- class(x)
>   dc <- if(.R.) c('POSIXt','POSIXct','dates','times','chron') else
>                 c('timeDate','date','dates','times','chron')
>   length(cl) && any(cl %in% dc)
> }
> 
> But describe also runs this command:
> 
>   if(isdatetime) notime <- all(format(x.unique,"%H%M%S")=='000000')
> 
> and such calls to format for chron objects (which actually calls
> format.dates if x.unique is a dates object) do not work.  So a few more
> modifications will need to be made in describe to work with chron objects,
> when time allows.
> 
> Frank
> 
> > 
> > Thank you!
> > 
> > Quoting Frank E Harrell Jr <feh3k at spamcop.net>:
> > 
> > > On Tue, 11 Nov 2003 18:28:24 -0500
> > > tmurph6 at po-box.mcgill.ca wrote:
> > > 
> > > > Hello,
> > > > 
> > > > I am trying to use HMisc describe on a data frame. I have specified
> > > > certain variables as dates using, for example:
> > > > 
> > > > df1$aidsdate <- dates(aidsdate,format="day.mon.year",
> > > > origin=c(month = 1, day = 1, year = 1960))
> > > > 
> > > > When I use describe on the dataframe I get this error:
> > > > 
> > > > Error in Ops.dates(weights, x) : * not defined for chron objects
> > > > 
> > > > Has anyone had this problem? I had used the same method with an
> > > > older version of R and HMisc and it worked. 
> > > > 
> > > > Should be be formatting date variables differently?
> > > > 
> > > > Thank you!
> > > > 
> > > > Sincerely,
> > > > Tanya Murphy
> > > > 
> > > 
> > > In recent versions, describe for R recognizes date/times by one of the
> > > following classes: 'POSIXt','POSIXct','chron'. If you really need to
> > > use another date format I could be talked into extending the code for
> > > that if you can remind me where dates( ) is found. -Frank
> > > 
> > > ---
> > > Frank E Harrell Jr Professor and Chair School of Medicine
> > > Department of Biostatistics Vanderbilt University
> > >
> > 
> > ______________________________________________
> > R-help at stat.math.ethz.ch mailing list
> > https://www.stat.math.ethz.ch/mailman/listinfo/r-help
> 
> 
> ---
> Frank E Harrell Jr    Professor and Chair            School of Medicine
>                       Department of Biostatistics    Vanderbilt University
> 




-------------------------------------------------
This mail sent through IMP: http://horde.org/imp/