[R] summarizing daily time-series date by month

[Benjamin M. Osborne]

Message: 63
Date: Wed, 26 Jan 2005 04:28:51 +0000 (UTC)
From: Gabor Grothendieck <ggrothendieck at myway.com>
Subject: Re: [R] chron: parsing dates into a data frame using a
        forloop
To: r-help at stat.math.ethz.ch
Message-ID: <loom.20050126T052153-333 at post.gmane.org>
Content-Type: text/plain; charset=us-ascii

Benjamin M. Osborne <Benjamin.Osborne <at> uvm.edu> writes:

:
: I have one data frame with a column of dates and I want to fill another data
: frame with one column of dates, one of years, one of months, one of a unique
: combination of year and month, and one of days, but R seems to have some
: problems with this.  My initial data frame looks like this (ignore the NAs in
: the other fields):
:
: > mans[1:10,]
:        date loc snow.new prcp tmin snow.dep tmax
: 1  11/01/54   2       NA   NA   NA       NA   NA
: 2  11/02/54   2       NA   NA   NA       NA   NA
: 3  11/03/54   2       NA   NA   NA       NA   NA
: 4  11/04/54   2       NA   NA   NA       NA   NA
: 5  11/05/54   2       NA   NA   NA       NA   NA
: 6  11/06/54   2       NA   NA   NA       NA   NA
: 7  11/07/54   2       NA   NA   NA       NA   NA
: 8  11/08/54   2       NA   NA   NA       NA   NA
: 9  11/09/54   2       NA   NA   NA       NA   NA
: 10 11/10/54   2       NA   NA   NA       NA   NA
: >
:
: The code and resultant data frame look like this:
:
: > for(i in 1:10){
: + mans.met$date[i]<-mans$date[i]
: + mans.met$year[i]<-years(mans.met$date[i])
: + mans.met$month[i]<-months(mans.met$date[i])
: + mans.met$yearmo[i]<-cut(mans.met$date[i], "months")
: + mans.met$day[i]<-days(mans.met$date[i])
: + }
: > mans.met[1:10,]
:        date year month yearmo day snow.new snow.dep prcp tmin tmax tmean
: 1  11/01/54    1    11      1   1       NA       NA   NA   NA   NA    NA
: 2  11/02/54    1    11      1   2       NA       NA   NA   NA   NA    NA
: 3  11/03/54    1    11      1   3       NA       NA   NA   NA   NA    NA
: 4  11/04/54    1    11      1   4       NA       NA   NA   NA   NA    NA
: 5  11/05/54    1    11      1   5       NA       NA   NA   NA   NA    NA
: 6  11/06/54    1    11      1   6       NA       NA   NA   NA   NA    NA
: 7  11/07/54    1    11      1   7       NA       NA   NA   NA   NA    NA
: 8  11/08/54    1    11      1   8       NA       NA   NA   NA   NA    NA
: 9  11/09/54    1    11      1   9       NA       NA   NA   NA   NA    NA
: 10 11/10/54    1    11      1  10       NA       NA   NA   NA   NA    NA
: >
:
: The problem seems to be with assigning within the forloop, or making the
: assignment into a data frame, since:
:
: > years(mans.met$date[5])
: [1] 1954
: Levels: 1954
: > test<-years(mans.met$date[5])
: > test
: [1] 1954
: Levels: 1954
: >
: > months(mans.met$date[5])
: [1] Nov
: 12 Levels: Jan < Feb < Mar < Apr < May < Jun < Jul < Aug < Sep < ... < Dec
: > test<-months(mans.met$date[5])
: > test
: [1] Nov
: 12 Levels: Jan < Feb < Mar < Apr < May < Jun < Jul < Aug < Sep < ... < Dec
: >
: > cut(mans.met$date[3], "months")
: [1] Nov 54
: Levels: Nov 54
: > test<-cut(mans.met$date[3], "months")
: > test
: [1] Nov 54
: Levels: Nov 54
: >
: > days(mans.met$date[4])
: [1] 4
: 31 Levels: 1 < 2 < 3 < 4 < 5 < 6 < 7 < 8 < 9 < 10 < 11 < 12 < 13 < ... < 31
: > test<-days(mans.met$date[4])
: > test
: [1] 4
: 31 Levels: 1 < 2 < 3 < 4 < 5 < 6 < 7 < 8 < 9 < 10 < 11 < 12 < 13 < ... < 31
: >
:
: Any suggestions will be appreciated.
: -Ben Osborne

I guess you set up mans.met as numeric columns and when you
assign your factors to numeric variables you get
the underlying codes.  Note that if f is a factor then as.numeric(f)
gives the codes underlying the factor whereas as.character(f) gives
the labels.

It would be better not to use a loop at all.  I don't know whether you
want or not want factors but at any rate here is something you could
try.  It creates data frame df2 without a loop.

df2 <- data.frame(date = mans$date, yearmo = as.character(cut(mans$date, "m")))
df2 <- cbind(df2, month.day.year(mans$date))

Finally, do you really want this redundant representation?  I would tend to
go with just storing the dates and computing any of the other quantities
on-the-fly as needed.

##########
The reason for the redundancy is that I will want to summarize these 50 years of
daily time series data by month, so that records that share each unique year
and month in the mans.met$yearmo column will be summed or averaged, etc. into a
new row in another data frame(mans.monthly, having
nrow=length(unique(mans.met$yearmo))).  The way I would do this is again using
a forloop, but the loop won't recognize :
     for (i in 1:(length(unique(mans.met$yearmo[i])))){

What I really need to know is why I can call any ith of
     unique(mans.met$yearmo[i])
by itself, but not in a loop.

Or, perhaps there is an even easier way to extract the year and month from the
date
column on the fly to compute these summaries?

Thanks,
Ben Osborne

-- 
Botany Department
University of Vermont
109 Carrigan Drive
Burlington, VT 05405

benjamin.osborne at uvm.edu
phone: 802-656-0297
fax: 802-656-0440