CODEWITHSUNDEEP
r
Thorst
Thorst

Reputation: 1610

R Using factor in a function

Im having some troubles using factors in functions, or just to make use of them in basic calculations. I have a data-frame something like this (but with as many as 6000 different factors). 

df<- data.frame( p <- runif(20)*100,
q = sample(1:100,20, replace = T),
tt = c("e","e","f","f","f","i","h","e","i","i","f","f","j","j","h","h","h","e","j","i"),
ta = c("a","a","a","b","b","b","a","a","c","c","a","b","a","a","c","c","b","a","c","b"))
colnames(df)<-c("p","q","ta","tt")

Now price = p and quantity = q are my variables, and tt and ta are different factors.

Now, I would first like to find the average price per unit of q by each different factor in tt

(p*q ) / sum(q) by tt

This would in this case give me a list of 3 different sums, by a, b and c (I have 6000 different factors so I need to do it smart :) ).

I have tried using split to make lists, and in this case i can get each individual tt factor to contain the prices and another for the quantity, but I cant seem to get them to for example make an average. I've also tried to use tapply, but again I can't see how I can incorporate factors into this?

EDIT: I can see I need to clearify:

I need to find 3 sums, the average price pr. q given each factor, so in this simplified case it would be:

a: Sum of p*q for (Row (1,2,3, 7, 11, 13,14,18) / sum (q for row Row (1,2,3, 7, 11, 13,14,18)

So the result should be the average price for a, b and c, which is just 3 values.

Upvotes: 1

Views: 280

Answers (2)

alap
alap

Reputation: 647

If I understood corectly you'r problem this should be the answer. Give it a try and responde, that I can adjust it if it's needed.

myRes <- function(tt) {

  out <- NULL;
  qsum <- sum(as.numeric(df[,"q"]))
  psum <- sum(as.numeric(df[,"p"]))
  for (var in tt) {
    index <- which(df["tt"] == var)

    out <- c(out, ((qsum *psum) / sum(df[index,"q"])))
  }
  return (out)
}

threeValue <- myRes(levels(df[, "tt"]));

Upvotes: 0

Paul Hiemstra
Paul Hiemstra

Reputation: 60984

I'd use plyr to do this:

library(plyr)
ddply(df, .(tt), mutate, new_col = (p*q) / sum(q))
          p  q ta tt     new_col
1  73.92499 70  e  a 11.29857879
2  58.49011 60  e  a  7.66245932
3  17.23246 27  f  a  1.01588711
4  64.74637 42  h  a  5.93743967
5  55.89372 45  e  a  5.49174103
6  25.87318 83  f  a  4.68880732
7  12.35469 23  j  a  0.62043207
8   1.19060 83  j  a  0.21576367
9  84.18467 25  e  a  4.59523322
10 73.59459 66  f  b 10.07726727
11 26.12099 99  f  b  5.36509998
12 25.63809 80  i  b  4.25528535
13 54.74334 90  f  b 10.22178577
14 69.45430 50  h  b  7.20480246
15 52.71006 97  i  b 10.60762667
16 17.78591 54  i  c  5.16365066
17  0.15036 41  i  c  0.03314388
18 85.57796 30  h  c 13.80289670
19 54.38938 44  h  c 12.86630433
20 44.50439 17  j  c  4.06760541

plyr does have a reputation for being slow, data.table provides similar functionality, but much higher performance.

Upvotes: 1

Related Questions