[R] Correctly applying aggregate.ts()

[Rui Barradas]

Hello,

Like Bert said, your data is a data.frame so there is no need to call 
aggregate.ts. Besides, R will call the right method so unless you want 
to change the standard behaviour, it would be enough to call aggregate 
and let the methods dispatch code to its job.

As for the problem, first an example of the formula interface, which I 
almost always prefer.


aggregate(prcp ~ substr(sampdate, 1, 7), data = dp, FUN = sum, na.rm = TRUE)
#  substr(sampdate, 1, 7) prcp
#1                2005-01 4.88
#2                2005-02 2.27
#3                2005-03 0.06


Now, you would have to change the name of the Month column, but it 
worked as expected, there was no NA issues.
And there is no need to subset the data.frame, R will find the columns 
where they are, by their names, as long as you pass the argument data = 
dp to aggregate.


If you want several statistics at the same time, it's a bit trickier, 
but with practice it becomes intuitive. (So to speak.)

Define a custom summary function. I haven't changed the default na.rm 
setting but it would make the rest of the code simpler to set na.rm = 
TRUE right now.

customSmry <- function(x, na.rm = FALSE){
   c(Sum = sum(x, na.rm = na.rm),
     Median = median(x, na.rm = na.rm),
     Max = max(x, na.rm = na.rm)
   )
}


#Now call aggregate:

agg <- aggregate(prcp ~ substr(sampdate, 1, 7), dp, FUN = customSmry, 
na.rm = TRUE)


But be VERY carefull, the result is not a df with 4 columns, it's a df 
with only two columns, the second being a matrix as you can see in the 
output of str.


str(agg)
#'data.frame':	3 obs. of  2 variables:
# $ substr(sampdate, 1, 7): chr  "2005-01" "2005-02" "2005-03"
# $ prcp                  : num [1:3, 1:3] 4.88 2.27 0.06 0.05 0 0.01 
1.12 0.65 0.05
# ..- attr(*, "dimnames")=List of 2
# .. ..$ : NULL
# .. ..$ : chr  "Sum" "Median" "Max"


So the final steps will be to cbind those two "columns" into a df.
"columns" is between quotes because I am not cbinding the first column, 
I'm cbinding the sub-df agg[1]. Like this the method of cbind that is 
called is cbind.data.frame and the result is a df.
Also, since df's are lists, the second column is an actual column but 
not a vector, an object of class matrix. This column is a list member, 
like all df columns and I will subset the df 'agg' as a list, agg[[2]].

As a bonus, the colnames of the matrix are immediately right, no prcp 
prefix. The first column's name comes from the function substr, and is 
not part of this story, just rename it when it's all done.


agg <- cbind(agg[1], agg[[2]])
str(agg)
#'data.frame':	3 obs. of  4 variables:
# $ substr(sampdate, 1, 7): chr  "2005-01" "2005-02" "2005-03"
# $ Sum                   : num  4.88 2.27 0.06
# $ Median                : num  0.05 0 0.01
# $ Max                   : num  1.12 0.65 0.05

names(agg)[1] <- "Month"
agg
#    Month  Sum Median  Max
#1 2005-01 4.88   0.05 1.12
#2 2005-02 2.27   0.00 0.65
#3 2005-03 0.06   0.01 0.05


Finally, try to get some practice with the formula interface, you will 
see that it pays in code simplicity and readability.


Hope this helps,

Rui Barradas

Às 22:19 de 07/09/2018, Rich Shepard escreveu:
>    I've read ?aggregate and several blog posts on using aggregate() yet I
> still haven't applied it correctly to my dataframe. The sample data are:
> 
> structure(list(sampdate = c("2005-01-01", "2005-01-02", "2005-01-03", 
> "2005-01-04", "2005-01-05", "2005-01-06", "2005-01-07", "2005-01-08", 
> "2005-01-09", "2005-01-10", "2005-01-11", "2005-01-12", "2005-01-13", 
> "2005-01-14", "2005-01-15", "2005-01-16", "2005-01-17", "2005-01-18", 
> "2005-01-19", "2005-01-20", "2005-01-21", "2005-01-22", "2005-01-23", 
> "2005-01-24", "2005-01-25", "2005-01-26", "2005-01-27", "2005-01-28", 
> "2005-01-29", "2005-01-30", "2005-01-31", "2005-02-01", "2005-02-02", 
> "2005-02-03", "2005-02-04", "2005-02-05", "2005-02-06", "2005-02-07", 
> "2005-02-08", "2005-02-09", "2005-02-10", "2005-02-11", "2005-02-12", 
> "2005-02-13", "2005-02-14", "2005-02-15", "2005-02-16", "2005-02-17", 
> "2005-02-18", "2005-02-19", "2005-02-20", "2005-02-21", "2005-02-22", 
> "2005-02-23", "2005-02-24", "2005-02-25", "2005-02-26", "2005-02-27", 
> "2005-02-28", "2005-03-01", "2005-03-02", "2005-03-03"), prcp = c(0.59, 
> 0.08, 0.1, 0, 0, 0.02, 0.05, 0.1, 0, 0.02, 0, 0.05, 0.2, 0, 0, 0.5, 
> 0.41, 0.84, 0.01, 0.1, 0.01, 0, 0, 0, 0, 0.21, 0.24, 0.13, 1.12, 0.01, 
> 0.09, 0, 0, 0, 0.35, 0.18, 0.65, 0.16, 0, 0, 0, 0, 0.55, 0.21, 0, 0, 0, 
> 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0.17, 0.05, 0.01, 0)), row.names = 
> c(NA, 62L), class = "data.frame")
> 
>    What I need to learn how to do is to calculate monthly sum, median, and
> maximum rainfall amounts from the full data set which has daily rainfall
> amounts. My most current effort to calculate monthly sums uses this syntax:
> 
> monthly.rain <- aggregate.ts(x = dp['sampdate','prcp'], by = list(month = \
> substr(dp$sampdate, 1, 7)), FUN = sum, na.rm = TRUE)
> 
> (entered on a single line) which produces this result:
> 
> head(monthly.rain)
> [1] NA
> 
>    The sample data has 62 of the 113K rows in the dataframe. A larger 
> set can
> be provided if needed.
> 
>    An explanation of what I've missed is needed.
> 
> Regards,
> 
> Rich
> 
> ______________________________________________
> R-help using r-project.org mailing list -- To UNSUBSCRIBE and more, see
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide 
> http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.