Rob

Thanks for the help. This works well enough.

For the audience, I think those who use multi-relational databases might
find my problem and solution helpful.

First, I want to highlight a silly error. When trying to create a dataframe
of the date vector, I was using "as.data.frame" and this wasn't working.
Using cbind creates the matrix and coerces the date vector to a number, so
even when I converted the "numerical" date to a "date" date, I couldn't get
it into the dataframe. I then checked previous code I wrote to convert to
dataframes and the correct code is simply "data.frame" instead of
"as.data.frame".

Here is my story in more detail.

I have one table with information about injuries, including injury date, but
I don't have the enddate of the injury in that file. I have another table
with treatment dates for every subject for every injury. If I can get the
last treatment date for each injury, I can the merge this with the Injury
file and I would have all the information in one file and be able to work
from there.

So, I split the file by injuryID. This creates a list of tables. Each
component of the list is a table that has all the fields in the original
injury file but each table is limited to only one injuryId.

One can then use "lapply" to apply a function to each item in the list (i.e.
to each table). In my case, I took the last treatment in each table by using
the "tail" function.

However, using this method, the function returns a list where the name of
each component is the InjuryId and the value of each component is the last
treatment date (the class of the components is indeed date). But to merge
the information to the Injury file, I need a vector of dates associated with
the InjuryId, not a "list" of dates. So, first, I created a vector of the
InjuryId from the list using "as.numeric(names(listname))". This effectively
converts the listnames (in my case, the InjuryId) to a numeric vector. Now
came the problem I asked about.

Now, I had to get the components of the list (i.e. the actual dates) into a
vector. I then learned about "unlist", which actually worked well and
converts the components in a "list" to a vector. However, it converts the
dates to numerical vectors, just as "cbind" does. Using Rob's trick, I
simply added the number to the date 1970-01-01 and now the vector is in date
class. To link it to the InjuryId, you simply have to use
"data.frame(InjuryId,Datefield)" and it works, keeping the date in date
format (this is where I had used "as.data.frame" or "cbind" and it just kept
converting everything back to a number.

Now, I have a data.frame with InjuryId and the Lasttreatment date, and
another file with the InjuryId and all the information. The last step is to
merge the files, and the help file was pretty easy to use for that feature.

Ian


>-----Original Message-----
>From: sportsci_rtutorial@yahoogroups.com
>[mailto:sportsci_rtutorial@yahoogroups.com]On Behalf Of gerrobrein
>Sent: 19-Aug-2007 4:45 PM
>To: sportsci_rtutorial@yahoogroups.com
>Subject: [sportsci_rtutorial] Re: New file uploaded to
>sportsci_rtutorial
>
>
>Hi Ian,
>
>I have to confess that I'm not too familiar with the Date class.
>
>Anyway, from what I've picked up from the help files, Date objects are
>indeed internally represented as integers which mark the days since
>1970-01-01.
>
>Sys.Date()
>unclass(Sys.Date())
>Sys.Date() - unclass(Sys.Date())
>
>Now, concerning cbind(), the help file reads:
>The type of a matrix result determined from the highest type of
>any of the inputs in the hierarchy:
> raw < logical < integer < real < complex < character < list
>Hence, there's no cbind() function for Date objects and the objects
>are coerced to numeric. That's why you end up with numbers in your
>matrix.
>
>I guess I didn't understand the problem properly, but can't you
>include the dates straight away into your date.frame()
>
>sDate <- seq.Date(as.Date("2007/08/01"),as.Date("2007/08/30"),"4
>days")
>sDate
>injuries <- 1:8
>DF <- data.frame(date=sDate, injuries=injuries)
>DF
>
>If you have to go with cbind() what you could do, although it's not
>all too elegant and more an ad-hoc solution, you could generate a
>Date() object representing the "first" day.
>
>origin <- as.Date("1970-01-01")
>
>Generate the date matrix with cbind()
>
>dMat <- cbind(sDate,sDate)
>dMat
>
>When you need you dates back, you extract them normally from dMat and
>add them to the origin object.
>
>origin + tab[3,2]
>
>This way you end up with a Date object again.
>I suppose you're running some functions over your matrix? You could
>include this operation into this function then.
>Hope that helps. If not maybe include some generated data which looks
>like the one you're working with.
>
>Cheers
>Rob
>
>--- In sportsci_rtutorial@yahoogroups.com, "Ian
>Shrier" <ian.shrier@...> wrote:
>>
>> Rob
>>
>> I have run into a problem with manipulating a date vector. When I
>try to
>> cbind a date vector, it converts the date to a numerical vector. I
>think
>> this is because cbind creates a matrix, and matrix has to have
>numbers if I
>> remember correctly.
>>
>> I have a friend who is really good with R and he showed me that this
>> represents the number of seconds since Jan 1, 1970. He was trying
>to convert
>> it with ISOdatetime and adding the number generated by cbind. This
>converts
>> it back to a date but we couldn't get it into a dataframe.
>>
>> To take a step back so this in context, i'm working with a multi-
>relational
>> database. One table has all the injuries of athletes with the date
>of
>> injury. The other table has all the treatment dates for each
>injury. I'm
>> using a function to look at the treatment table and pull out the
>last
>> treatment date for each injury and I want to put this date into the
>> dataframe that has the list of injuries. This would give me one
>table for
>> injuries that includes the last treatment date as part of the
>table. The
>> date is then used to calculate number of days for the injury to
>heal, and
>> also will be used as the end-date when searching through a third
>table that
>> has exposures so I can count how many exposures from the injury
>start date
>> to end date.
>>
>> Any ideas?
>>
>> Ian
>>
>
>
>
>
>
>Yahoo! Groups Links
>
>
>