r- mean over variables using ddply

Question

I have a table (t02f) with a character variable ("t02.clase") and numeric variables. I need to create another table with the mean of each variable for each value in the column t02.clase.

Example table:

t02.clase Std_A_CLI_monto_sucursal_1 Std_A_CLI_monto_sucursal_2 Std_A_CLI_monto_sucursal_3
1        AK                          0                 0.00000000                0.051311597
2        AK                          0                 0.01303586                0.003442244
3        AK                          0                 0.00000000                0.017347593
  Std_A_CLI_monto_sucursal_4 Std_A_CLI_promociones Std_A_CLI_dias_cliente Std_A_CLI_sucursales
1                          0             0.4736842             0.57061341           0.05555556
2                          0             0.5000000             0.55492154           0.05555556
3                          0             0.0000000             0.05991441           0.05555556
  Std_A_CLI_promos_enviadas Std_A_CLI_promos_compras Std_A_CLI_produni Std_A_CLI_flagret
1                 0.6363636                0.2727273        0.14749987                 0
2                 0.6363636                0.1818182        0.05973218                 1

I need to end up with sth like:

t02.clase   Std_A_CLI_monto_sucursal_1  Std_A_CLI_monto_sucursal_2  Std_A_CLI_monto_sucursal_3
AK               0.3                                0.001                       0.2
AM               0.2                                0.02                        0.3
ES               0.5                                0.001                       0.0001

I used:

a1<-data.frame(NULL)
a<-data.frame(NULL)
for (i in (colnames(t02f)[-1])) {a1<-ddply(t02f, .(t02.clase), summarize, mean(i))      
                          a<-cbind(a1,a) }

But I get numbers from 1 to ncol(t02f) insted of the means.

Answer 1

Or using data.table

library(data.table)
setDT(t02f)[, lapply(.SD, mean, na.rm = TRUE), by = t02.clase]
#    t02.clase Std_A_CLI_monto_sucursal_1 Std_A_CLI_monto_sucursal_2 Std_A_CLI_monto_sucursal_3 Std_A_CLI_monto_sucursal_4 Std_A_CLI_promociones Std_A_CLI_dias_cliente
# 1:        AK                          0                0.004345287                 0.02403381                          0             0.3245614              0.3951498
#    Std_A_CLI_sucursales
# 1:           0.05555556

---

Frankly, you could do it quite efficiently using base R by and colMeans combination, something like (don't run it if already converted your data set to data.table class)

do.call(rbind, by(t02f[, -1], t02f$t02.clase, colMeans,  na.rm = TRUE))
#    Std_A_CLI_monto_sucursal_1 Std_A_CLI_monto_sucursal_2 Std_A_CLI_monto_sucursal_3 Std_A_CLI_monto_sucursal_4 Std_A_CLI_promociones Std_A_CLI_dias_cliente
# AK                          0                0.004345287                 0.02403381                          0             0.3245614              0.3951498
#    Std_A_CLI_sucursales
# AK           0.05555556

Or (just for fun)

vapply(t02f[, -1], function(x) tapply(x,  t02f$t02.clase, mean, rm = TRUE), double(1))

Answer 2

Try summarise_each from dplyr

library(dplyr)
res <- t02f %>%
            group_by(t02.clase)%>%
            summarise_each(funs(mean=mean(., na.rm=TRUE)))  


as.data.frame(res)
#  t02.clase Std_A_CLI_monto_sucursal_1 Std_A_CLI_monto_sucursal_2
#1        AK                          0                0.004345287
#   Std_A_CLI_monto_sucursal_3 Std_A_CLI_monto_sucursal_4 Std_A_CLI_promociones
#1                 0.02403381                          0             0.3245614
#  Std_A_CLI_dias_cliente Std_A_CLI_sucursales
#1              0.3951498           0.05555556

If you want to use ddply , there is an option colwise

 library(plyr)
 ddply(t02f, .(t02.clase), numcolwise(mean, na.rm=TRUE))
 #  t02.clase Std_A_CLI_monto_sucursal_1 Std_A_CLI_monto_sucursal_2
 #1        AK                          0                0.004345287
 #  Std_A_CLI_monto_sucursal_3 Std_A_CLI_monto_sucursal_4 Std_A_CLI_promociones
 #1                 0.02403381                          0             0.3245614
 #  Std_A_CLI_dias_cliente Std_A_CLI_sucursales
 #1              0.3951498           0.05555556

Or using aggregate from base R

 aggregate(.~t02.clase, t02f, mean, na.rm=TRUE)

data

t02f <- structure(list(t02.clase = c("AK", "AK", "AK"),
 Std_A_CLI_monto_sucursal_1 = c(0L,0L, 0L), Std_A_CLI_monto_sucursal_2 = 
 c(0, 0.01303586, 0), Std_A_CLI_monto_sucursal_3 = c(0.051311597, 
 0.003442244, 0.017347593), Std_A_CLI_monto_sucursal_4 = c(0L, 
 0L, 0L), Std_A_CLI_promociones = c(0.4736842, 0.5, 0), Std_A_CLI_dias_cliente =
 c(0.57061341,0.55492154, 0.05991441), Std_A_CLI_sucursales = c(0.05555556, 
 0.05555556, 0.05555556)), .Names = c("t02.clase", "Std_A_CLI_monto_sucursal_1", 
 "Std_A_CLI_monto_sucursal_2", "Std_A_CLI_monto_sucursal_3",
"Std_A_CLI_monto_sucursal_4","Std_A_CLI_promociones", "Std_A_CLI_dias_cliente",
"Std_A_CLI_sucursales"), row.names = c("1", "2", "3"), class = "data.frame")