[R] weighted average grouped by variables

[Rui Barradas]

Hello,

Using base R only, the following seems to do what you want.

with(mydf, ave(speed, date_time, type, FUN = weighted.mean, w = n_vehicles))


Hope this helps,

Rui Barradas

Em 09-11-2017 13:16, Massimo Bressan escreveu:
> Hello
>
> an update about my question: I worked out the following solution (with the package "dplyr")
>
> library(dplyr)
>
> mydf%>%
> mutate(speed_vehicles=n_vehicles*mydf$speed) %>%
> group_by(date_time,type) %>%
> summarise(
> sum_n_times_speed=sum(speed_vehicles),
> n_vehicles=sum(n_vehicles),
> vel=sum(speed_vehicles)/sum(n_vehicles)
> )
>
>
> In fact I was hoping to manage everything in a "one-go": i.e. without the need to create the "intermediate" variable called "speed_vehicles" and with the use of the function weighted.mean()
>
> any hints for a different approach much appreciated
>
> thanks
>
>
>
> Da: "Massimo Bressan" <massimo.bressan at arpa.veneto.it>
> A: "r-help" <r-help at r-project.org>
> Inviato: Giovedì, 9 novembre 2017 12:20:52
> Oggetto: weighted average grouped by variables
>
> hi all
>
> I have this dataframe (created as a reproducible example)
>
> mydf<-structure(list(date_time = structure(c(1508238000, 1508238000, 1508238000, 1508238000, 1508238000, 1508238000, 1508238000), class = c("POSIXct", "POSIXt"), tzone = ""),
> direction = structure(c(1L, 1L, 1L, 1L, 2L, 2L, 2L), .Label = c("A", "B"), class = "factor"),
> type = structure(c(1L, 2L, 3L, 4L, 1L, 2L, 3L), .Label = c("car", "light_duty", "heavy_duty", "motorcycle"), class = "factor"),
> avg_speed = c(41.1029082774049, 40.3333333333333, 40.3157894736842, 36.0869565217391, 33.4065155807365, 37.6222222222222, 35.5),
> n_vehicles = c(447L, 24L, 19L, 23L, 706L, 45L, 26L)),
> .Names = c("date_time", "direction", "type", "speed", "n_vehicles"),
> row.names = c(NA, -7L),
> class = "data.frame")
>
> mydf
>
> and I need to get to this final result
>
> mydf_final<-structure(list(date_time = structure(c(1508238000, 1508238000, 1508238000, 1508238000), class = c("POSIXct", "POSIXt"), tzone = ""),
> type = structure(c(1L, 2L, 3L, 4L), .Label = c("car", "light_duty", "heavy_duty", "motorcycle"), class = "factor"),
> weighted_avg_speed = c(36.39029, 38.56521, 37.53333, 36.08696),
> n_vehicles = c(1153L,69L,45L,23L)),
> .Names = c("date_time", "type", "weighted_avg_speed", "n_vehicles"),
> row.names = c(NA, -4L),
> class = "data.frame")
>
> mydf_final
>
>
> my question:
> how to compute a weighted mean i.e. "weighted_avg_speed"
> from "speed" (the values whose weighted mean is to be computed) and "n_vehicles" (the weights)
> grouped by "date_time" and "type"?
>
> to be noted the complication of the case "motorcycle" (not present in both directions)
>
> any help for that?
>
> thank you
>
> max
>
>
>