R help - Jan 2012 - For Loop Error

Hi Again,

I am writing a 'for loop' to create a matrix of randomly sampled colors.
I've written this loop in matlab and it works fine.  I then tried to do it
in R and apparently there is something wrong with my syntax b/c every time I
run the script, the for loop "blows up" at a different point in the
loop.

As you can see, I ask R to clear my workspace each time, so there shouldn't
be any variables in my workspace that are messing this up.

rm(list=ls())

# Sample Size
size <- 53  
# Probability of each color based on company
P.company <- c(.14,.14,.16,.13,.20,.24)
# Names of colors
color <-
c('Br','Y','G','R','O','Bl')
# Make an empty matrix that will be filled in by for loop
table.combos <- matrix(nrow = 10000, ncol = 6);

# For loop will run through choosing a random bag of 53 M&Ms 10000 times
# and create a table enumerating the number of each color in each of these
10000 bags
for(i in 1:10000) {
  combos <- sample(color,size, replace = TRUE, prob = P.company)
  table.combos[i, ] <- table(combos)
 
colnames(table.combos)<-c("Br","Y","G","R","O","Bl")
}

Every time the loop blows up, I get back this error:

Error in table.combos[i, ] <- table(combos) : 
  number of items to replace is not a multiple of replacement length

There is no apparent consistency on which "i" in the loop I get this
error.
Sometimes i = 10, sometimes i = 685, sometimes i = 1954, sometimes i = 59,
etc.

If anyone can please tell me what I am doing wrong, it would be greatly
appreciated!

Cheers,
Melissa

--
View this message in context:
http://r.789695.n4.nabble.com/For-Loop-Error-tp4336585p4336585.html
Sent from the R help mailing list archive at Nabble.com.

Hello,
> 
> Every time the loop blows up, I get back this error:
> 
> Error in table.combos[i, ] <- table(combos) :
>   number of items to replace is not a multiple of replacement length
> 
This is because not all colors are present in that one sample that breaks
the code.
> 
> There is no apparent consistency on which "i" in the loop I get
this
> error.  
> 
There shouldn't be, every time you run the loop, 'sample' will take
different values. To get
reproducibility you would need 'set.seed'.

A solution could be based on the following.
It's important to sort, like the difference between 'tc1' and
'tc2' proves.
It's 'tc2' you want.

Note that I've reduced the sample size to 10.
Note also that I initialize the matrices to zeros, not NA's.


color <-
sort(c('Br','Y','G','R','O','Bl'))

tc1 <- matrix(0, nrow = 1, ncol = 6)
tc2 <- matrix(0, nrow = 1, ncol = 6)

colnames(tc1) <- color
colnames(tc2) <- color

set.seed(123)
for(i in 1) {
  combos <- sample(color, 10, replace = TRUE, prob = P.company)
  tc1[i, unique(combos)] <- table(combos)
  tc2[i, sort(unique(combos))] <- table(combos)
}

table(combos)
tc1
tc2


I hope this helps.

Rui Barradas


--
View this message in context:
http://r.789695.n4.nabble.com/For-Loop-Error-tp4336585p4337469.html
Sent from the R help mailing list archive at Nabble.com.

On Sat, Jan 28, 2012 at 10:25:47AM -0800, Melrose2012
wrote:
> Hi Again,
> 
> I am writing a 'for loop' to create a matrix of randomly sampled
colors.
> I've written this loop in matlab and it works fine.  I then tried to do
it
> in R and apparently there is something wrong with my syntax b/c every time
I
> run the script, the for loop "blows up" at a different point in
the loop.
> 
> As you can see, I ask R to clear my workspace each time, so there
shouldn't
> be any variables in my workspace that are messing this up.
> 
> rm(list=ls())
> 
> # Sample Size
> size <- 53  
> # Probability of each color based on company
> P.company <- c(.14,.14,.16,.13,.20,.24)
> # Names of colors
> color <-
c('Br','Y','G','R','O','Bl')
> # Make an empty matrix that will be filled in by for loop
> table.combos <- matrix(nrow = 10000, ncol = 6);
> 
> # For loop will run through choosing a random bag of 53 M&Ms 10000
times
> # and create a table enumerating the number of each color in each of these
> 10000 bags
> for(i in 1:10000) {
>   combos <- sample(color,size, replace = TRUE, prob = P.company)
>   table.combos[i, ] <- table(combos)
>  
colnames(table.combos)<-c("Br","Y","G","R","O","Bl")
> }
> 
> Every time the loop blows up, I get back this error:
> 
> Error in table.combos[i, ] <- table(combos) : 
>   number of items to replace is not a multiple of replacement length
Hi.

Rui Barradas already pointed out that the problem is with samples,
which do not contain all colors. Forcing table() to produce zero
frequencies of colors, which do not occur, a factor may be used.
Try, for example

  size <- 53
  P.company <- c(.14,.14,.16,.13,.20,.24)
  color <-
c('Br','Y','G','R','O','Bl')
  table.combos <- matrix(nrow = 10000, ncol = 6);
  colnames(table.combos) <- color
 
  for(i in 1:10000) {
    combos <- sample(color,size, replace = TRUE, prob = P.company)
    table.combos[i, ] <- table(factor(combos, levels=color))
  }

  which(table.combos==0, arr.ind=TRUE)

         row col
   [1,]  588   1
   [2,] 3005   1
   [3,] 4535   1
   [4,] 8314   1
   [5,] 7654   2
   [6,] 8607   3
   [7,] 1790   4
   [8,] 1980   4
   [9,] 2991   4
  [10,] 4550   4
  [11,] 5868   4
  [12,] 6704   4
  [13,] 8769   4
  [14,] 8823   4
  [15,] 7861   5

Hope this helps.

Petr Savicky.