R help - Jul 2009 - Working with tables with missing levels

Hello

I'm trying to write a function to calculate the relative entropy between
two distributions. The data I have is in table format, for example:
> t1 <- prop.table(table(c(0,0,2,4,4)))
> t2 <- prop.table(table(c(0,2,2,2,3)))
> t1
  0   2   4 
0.4 0.2 0.4 
> t2
  0   2   3 
0.2 0.6 0.2

The relative entropy is given by

  H[P||Q] = sum(p * log2(p/q))

with the conventions that 0*log2(0/q) = 0 and p*log2(p/0) = Inf.

I'm not sure about what is the best way to achieve that. Is there a way
to test if a table has a value for a given level, so that I can detect
that, for example, t1 is missing levels 1 and 3 and t2 is missing levels
1 and 4 (is "level" the correct terminology here?)? Simply trying to
access t1[["1"]], for example, gives a "subscript out of
bounds" error.

Another option would be to "expand" the tables, so that, for example,
t1
becomes

  0   1   2   3   4 
0.4 0.0 0.2 0.0 0.4

Is there a way to do that?

Thanks,
Andre

Hi Andre,
Just about expending the table,

The way you could do this is by using factors, for example:
t1 <- prop.table(table(factor(c(0,0,2,4,4))))
t2 <- prop.table(table(factor( c(0,2,2,2,3))))

The rest is for more knowledgeable people then me to say...





On Mon, Jul 27, 2009 at 10:21 PM, Andre Nathan
<andre@digirati.com.br>wrote:
> Hello
>
> I'm trying to write a function to calculate the relative entropy
between
> two distributions. The data I have is in table format, for example:
>
> > t1 <- prop.table(table(c(0,0,2,4,4)))
> > t2 <- prop.table(table(c(0,2,2,2,3)))
> > t1
>
>  0   2   4
> 0.4 0.2 0.4
> > t2
>
>  0   2   3
> 0.2 0.6 0.2
>
> The relative entropy is given by
>
>  H[P||Q] = sum(p * log2(p/q))
>
> with the conventions that 0*log2(0/q) = 0 and p*log2(p/0) = Inf.
>
> I'm not sure about what is the best way to achieve that. Is there a way
> to test if a table has a value for a given level, so that I can detect
> that, for example, t1 is missing levels 1 and 3 and t2 is missing levels
> 1 and 4 (is "level" the correct terminology here?)? Simply trying
to
> access t1[["1"]], for example, gives a "subscript out of
bounds" error.
>
> Another option would be to "expand" the tables, so that, for
example, t1
> becomes
>
>  0   1   2   3   4
> 0.4 0.0 0.2 0.0 0.4
>
> Is there a way to do that?
>
> Thanks,
> Andre
>
> ______________________________________________
> R-help@r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide
> http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.
>


-- 
----------------------------------------------


My contact information:
Tal Galili
Phone number: 972-50-3373767
FaceBook: Tal Galili
My Blogs:
http://www.r-statistics.com/
http://www.talgalili.com
http://www.biostatistics.co.il

	[[alternative HTML version deleted]]

Try this:

t1 <- prop.table(table(factor(c(0,0,2,4,4), levels = 0:4)))
t2 <- prop.table(table(factor(c(0,2,2,2,3), levels = 0:4)))

On Mon, Jul 27, 2009 at 4:21 PM, Andre Nathan <andre@digirati.com.br>
wrote:
> Hello
>
> I'm trying to write a function to calculate the relative entropy
between
> two distributions. The data I have is in table format, for example:
>
> > t1 <- prop.table(table(c(0,0,2,4,4)))
> > t2 <- prop.table(table(c(0,2,2,2,3)))
> > t1
>
>  0   2   4
> 0.4 0.2 0.4
> > t2
>
>  0   2   3
> 0.2 0.6 0.2
>
> The relative entropy is given by
>
>  H[P||Q] = sum(p * log2(p/q))
>
> with the conventions that 0*log2(0/q) = 0 and p*log2(p/0) = Inf.
>
> I'm not sure about what is the best way to achieve that. Is there a way
> to test if a table has a value for a given level, so that I can detect
> that, for example, t1 is missing levels 1 and 3 and t2 is missing levels
> 1 and 4 (is "level" the correct terminology here?)? Simply trying
to
> access t1[["1"]], for example, gives a "subscript out of
bounds" error.
>
> Another option would be to "expand" the tables, so that, for
example, t1
> becomes
>
>  0   1   2   3   4
> 0.4 0.0 0.2 0.0 0.4
>
> Is there a way to do that?
>
> Thanks,
> Andre
>
> ______________________________________________
> R-help@r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide
> http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.
>


-- 
Henrique Dallazuanna
Curitiba-Paraná-Brasil
25° 25' 40" S 49° 16' 22" O

	[[alternative HTML version deleted]]

On Mon, 2009-07-27 at 16:34 -0300, Henrique Dallazuanna
wrote:
> Try this:
> 
> t1 <- prop.table(table(factor(c(0,0,2,4,4), levels = 0:4)))
> t2 <- prop.table(table(factor(c(0,2,2,2,3), levels = 0:4)))
Is there a way to do this given an already existing table? The problem
is that I actually build the distributions as I read data from files,
something like

  distr <- NULL
  for (file in files) {
    x <- as.matrix(read.table(file))
    t <- c(distr, table(x))
    distr <- tapply(t, names(t), sum)
  }
  distr <- prop.table(distr)

So I only know the maximum level after the distributions are created.

Thanks,
Andre