Aggregate df based on columns and group result

Question

I am trying to do the following, my dataset looks like this it contains a date in POSIXct format, hourly windspeed and hourly wind direction (df is called wind_DNSeason). My goal is to get frequency counts of windspeed according to the beaufort scale based on season and daylight.

  date                     wspd_havg10m_kn avg_wdir
1 2013-12-06 00:25:00        9.835853       50
2 2013-12-06 01:25:00       10.506479       56
3 2013-12-06 02:25:00       11.847732       55
4 2013-12-06 03:25:00        8.494600       53
5 2013-12-06 04:25:00       13.188985       47
6 2013-12-06 05:25:00       13.188985       60

Adding season based on the date:

wind_DNSeason$season<-time2season(wind_DNSeason$date, out.fmt="seasons", type="default")

Then I am cutting the data into daylight and nighttime using the openair package:

wind_DNSeason$daylight <- cutData(wind, type = "daylight", local.hour.offset = -8, latitude = 54.312519, longitude = -130.305405, local.tz= "Canada/Pacific")

I am aware of the function aggregate but I doubt I am using it correctly:

aggregate(wspd_havg10m_kn ~ season + daylight, wind_DNSeason, length)

That gives me the count of occurences but that is not what I want. Am I trying to do too much in one step?

I would need to know the grouping of the occuring windspeeds (see breaks below) per season split up in day and night. As I would like to create barplots with the different frequencies. breaks=c(0,1,3,6,10,16, 21, 27, 33, 40, 47)

Could I get something that would look somehow like this from which I could then easily calculate the percentages to plot it in barplots:

  season  daylight            total_count  wspd<=1 wspd>1,<=3 wspd>3,<=6 etc

1 autumm  daylight             854            151      34         56   
2 spring  daylight            2580            456      56         98
3 summer  daylight            1722            34       344        09
4 winter  daylight             852            545      55         55
5 autumm nighttime            1030            55        6         777
6 spring nighttime            1825            89       89         344
7 summer nighttime             827            344      55         66
8 winter nighttime            1533            34       66         777

any ideas? THanks for any help!

I tried using dplyr and I think I am really close but somehow it doesn't seem to add up the frequencies correctly. This is how I applied the suggested code:

a<-wind_DNSeason %>% group_by(season,daylight) %>% 
  mutate(count=n(),"wspd<=1" = sum(wspd_havg10m_kn<=1),
     "wspd>1,<=3" = sum(wspd_havg10m_kn > 1 & wspd_havg10m_kn <= 3, na.rm=TRUE), 
     "wspd>3,<=6" = sum(wspd_havg10m_kn > 3 & wspd_havg10m_kn <= 6,na.rm=TRUE),
     "wspd>6,<=10" = sum(wspd_havg10m_kn > 6 & wspd_havg10m_kn <= 10,na.rm=TRUE),
     "wspd>10,<=16" = sum(wspd_havg10m_kn > 10 & wspd_havg10m_kn <= 16,na.rm=TRUE),
     "wspd>16,<=21" = sum(wspd_havg10m_kn > 16 & wspd_havg10m_kn <= 21,na.rm=TRUE),
     "wspd>21,<=27" = sum(wspd_havg10m_kn > 21 & wspd_havg10m_kn <= 27,na.rm=TRUE),
     "wspd>27,<=33" = sum(wspd_havg10m_kn > 27 & wspd_havg10m_kn <= 33,na.rm=TRUE),
     "wspd>33,<=40" = sum(wspd_havg10m_kn > 33 & wspd_havg10m_kn <= 40,na.rm=TRUE),
     "wspd>40,<=47" = sum(wspd_havg10m_kn > 33 & wspd_havg10m_kn <= 47,na.rm=TRUE))

And the output looks like this, I selected some of the unique rows as it duplicates it across the whole df (e.g for winter day and nightime):

date    wspd_havg10m_kn avg_wdir    daylight    season  count   wspd<=1 wspd>1,<=3  wspd>3,<=6  wspd>6,<=10 wspd>10,<=16    wspd>16,<=21    wspd>21,<=27    wspd>27,<=33    wspd>33,<=40    wspd>40,<=47
1   2013-12-06 00:25:00 9.8358531   50  nighttime   winter  2751    NA  59  185 315 551 260 106 47  6   6
2   2013-12-06 12:25:00 7.3768898   57  daylight    winter  1449    NA  13  73  251 322 133 46  13  0   0

Shouldn't the frequencies of the different groups add up to the total count? The total df contains 13368 timesteps, if I add up the frequencies for each group I only get 11165. There are no windspeeds that are bigger than the largest group. What am I missing?

Answer 1

Here's a dplyr solution:

library(dplyr)
wind_DNSeason %>% group_by(season,daylight) %>% 
    summarise(count=n(),"wspd<=1" = sum(wspd_havg10m_kn<=1),
           "wspd>1,<=3" = sum(wspd_havg10m_kn > 1 & wspd_havg10m_kn <= 3),
           "wspd>3,<=6" = sum(wspd_havg10m_kn > 3 & wspd_havg10m_kn <= 6)
    )

You can add on as many columns for wind strengths as you want, filling out the names and requirements.

Answer 2

You mention plyr in your comments, so you can do this with:

library("plyr")

ddply(wind_DNSeason, .(season, daylight), summarize, n = length(wspd_havg10m_kn),
     "wspd<=1" = sum(wspd_havg10m_kn <= 1))

Additionally, if you want to automate the creation of those calculated values, you can do:

calc = function(x) {
   cuts = c(1, 3, 6, 10)
   res = data.frame(n = nrow(x))
   for(i in 1:(length(cuts) - 1)) {
       nm = sprintf("wspd>%d, <=%d", cuts[i], cuts[i + 1])
       val = sum(x$wspd_havg10m_kn > cuts[i] & x$wspd_havg10m_kn < cuts[i + 1], na.rm = T)
       res[, nm] = val
   }
   return(res)
}

ddply(wind_DNSeason, .(season, daylight), "calc")