Skip to content

subtract values

2 messages · Kara Przeczek, David Winsemius

Kara Przeczek
# 
Dear R-help
I am using R version 2.6.2. I am trying to subtract specific values from a larger data frame. I feel this should be straightforward, but I am struggling.
I have a dataframe "Bk" as follows:
DateTime	 cumPrecip	
01/01/2008 00:00	 348	
01/01/2008 01:00	 348	
01/01/2008 02:00	 348	
01/01/2008 03:00	 347	
01/01/2008 04:00	 348	
01/01/2008 05:00	 348	
01/01/2008 06:00	 349	
01/01/2008 07:00	 349	
01/01/2008 08:00	 349	
01/01/2008 09:00	 348	
01/01/2008 10:00	 349	
...                                    ...
 
I would like to subtract the cumulative precipitation value at 01/01/2008 00:00 from the value at 01/02/2008 00:00 and so on. Thus, I would like to subtract cumPrecip at DateTime i from DateTime i+23, for each 24 hour period.
I tried using for loop:
dPpt=0
for (i in 1:length(Bk[,"cumPrecip"]))
{dPpt[i]=diff(Bk[i,(i+23),"cumPrecip"])}
dPpt=dPpt[seq(1,length(dPpt),24)]
 
But it had many errors. I also tried using diff(Bk$cumPrecip, lag=23) but this moves through the data one step at a time and thus calculates a value for every hour of each day, not just midnight.
 
Thank you very much for your time and assistance!
Kara
 
 
 
Kara Przeczek
M.Sc. Candidate  NRES - Environmental Science
University of Northern B.C.
3333 University Way
Prince George B.C   V2N 4Z9
Phone: (250) 960-5427
przeczek at unbc.ca
David Winsemius
# 
I think the subject heading should be "looping one day at a time". I  
am guessing that one of the zoo functions has already solved that  
problem but it's not a package that I have much (or any) depth. If you  
create a date vector and pass it through as.POSIXct(), you will get a  
series of "midnights".

 >dts <- dates(c("02/27/92", "02/27/92", "01/14/92",
+                "02/28/92", "02/01/92"))
 > dts+2
[1] 02/29/92 02/29/92 01/16/92 03/01/92 02/03/92
 > as.POSIXct(dts)
[1] "1992-02-26 19:00:00 EST" "1992-02-26 19:00:00 EST" "1992-01-13  
19:00:00 EST" "1992-02-27 19:00:00 EST"
[5] "1992-01-31 19:00:00 EST"

Whether that will match what you are calling Bk$DateTime may depend on  
the actual type of those values, which you haven't told us. I do not  
think the call to the for-loop ought to be length of cumPrecip,  
either. Perhaps floor(length(Bk$cumPrecip)/24) or maybe just look at  
the data and figure out what the max of the dates would be.  Perhaps:
 > startdate <- dts[1]
 > ndays < -5
 > for (i in as.POSIXct(startdate +0:ndays) ) {print(as.POSIXct(i,  
origin="1970-01-01"))}
[1] "1992-02-27 EST"
[1] "1992-02-28 EST"
[1] "1992-02-29 EST"
[1] "1992-03-01 EST"
[1] "1992-03-02 EST"
[1] "1992-03-03 EST"