Caros,

rotineiramente me deparo com um problema que não sei como resolver. Considerem o seguinte exemplo:

(c1 <- trunc(runif(1000,1,10)))

(c2 <- trunc(runif(1000,5,70)))

(c3 <- trunc(runif(1000,1,4)))

(var <- rnorm(1000,10,2))

(peso <- round(rnorm(1000,1,.2)^2,1))

df <- data.frame(c1,c2,c3,var,peso)

Gostaria de uma função que pudesse me fornecer em uma única tabela do tipo

cols = c1,c2,c3,count,medvar

e, mais genericamente,

cols = c1,c2,c3,somapeso,medpondvar

com todas as combinações existentes (as vezes todas as teoricamente possiveis) entre c1,c2,c3

c1*c2

c1*c3

c2*c3

c1*c2*c3

De forma mais específica, gostaria:

t1 <- aggregate(list(count = df$var), list(c1=rep('*tot',1000) ,c2=rep('*tot',1000),c3=rep('*tot',1000)),length)

tc1 <- aggregate(list(count = df$var), list(c1=df$c1 ,c2=rep('*tot',1000),c3=rep('*tot',1000)),length)

tc2 <- aggregate(list(count = df$var), list(c1=rep('*tot',1000),c2=df$c2 ,c3=rep('*tot',1000)),length)

tc3 <- aggregate(list(count = df$var), list(c1=rep('*tot',1000),c2=rep('*tot',1000),c3=df$c3 ),length)

tc1c2 <- aggregate(list(count = df$var), list(c1=df$c1 ,c2=df$c2 ,c3=rep('*tot',1000)),length)

tc1c3 <- aggregate(list(count = df$var), list(c1=df$c1 ,c2=rep('*tot',1000),c3=df$c3 ),length)

tc2c3 <- aggregate(list(count = df$var), list(c1=rep('*tot',1000),c2=df$c2 ,c3=df$c3 ),length)

tc1c2c3 <- aggregate(list(count = df$var), list(c1=df$c1 ,c2=df$c2 ,c3=df$c3),length)

comp1 <- rbind(t1,tc1,tc2,tc3,tc1c2,tc1c3,tc2c3,tc1c2c3)

rm(t1,tc1,tc2,tc3,tc1c2,tc1c3,tc2c3,tc1c2c3)

t1 <- aggregate(list(medvar = df$var), list(c1=rep('*tot',1000) ,c2=rep('*tot',1000),c3=rep('*tot',1000)),mean)

tc1 <- aggregate(list(medvar = df$var), list(c1=df$c1 ,c2=rep('*tot',1000),c3=rep('*tot',1000)),mean)

tc2 <- aggregate(list(medvar = df$var), list(c1=rep('*tot',1000),c2=df$c2 ,c3=rep('*tot',1000)),mean)

tc3 <- aggregate(list(medvar = df$var), list(c1=rep('*tot',1000),c2=rep('*tot',1000),c3=df$c3 ),mean)

tc1c2 <- aggregate(list(medvar = df$var), list(c1=df$c1 ,c2=df$c2 ,c3=rep('*tot',1000)),mean)

tc1c3 <- aggregate(list(medvar = df$var), list(c1=df$c1 ,c2=rep('*tot',1000),c3=df$c3 ),mean)

tc2c3 <- aggregate(list(medvar = df$var), list(c1=rep('*tot',1000),c2=df$c2 ,c3=df$c3 ),mean)

tc1c2c3 <- aggregate(list(medvar = df$var), list(c1=df$c1 ,c2=df$c2 ,c3=df$c3),mean)

comp2 <- rbind(t1,tc1,tc2,tc3,tc1c2,tc1c3,tc2c3,tc1c2c3)

rm(tc1,tc2,tc3,tc1c2,tc1c3,tc2c3,tc1c2c3)

comp <- merge(comp1,comp2,by=c('c1','c2','c3'),all=T)

rm(comp1,comp2)

Exite algum pacote que faca isso?