boxplot

I have a matrix of occurences (7 rows(countries),46 columns(number of shoes)), n - number of occurences - for example in "a" I have n1=450, single pair of shoes...

   1   2   3  .... 46
a n1  n2   n3 .... n46

b

.

. . g

I would like to produce 7 boxplots representing the distribution in each category a-g. Normally I would have 450 times 1, n2 times 2 etc... is there a simple way to do it? Does anybody have a suggestion for a nicer representation than the boxplot?

Thank you

Answer 1

[Tom Lane]

This may partially answer what you want. Step 1 would be to take the data for "a" and stretch it out as you describe. Here's one way to expand a data vector according to a vector of frequencies:

f = [4 3 0 1 2];    % frequencies
d = [1 2 3 4 5];    % data
c = [0, cumsum(f)]; % cumulative sum of frequencies
x = zeros(1,c(end));
for j=1:length(c)-1
   x(c(j)+1:c(j+1)) = d(j); % insert each value as required
end

[New]

Thank you!

[Benoit Espinola]

This could become an issue if you have very large frequencies.

For instance:

d = [1 2 3 4 5];
f = [4e10 3e9 0 1e10 2e11];

You could endup with a variable with a length of 2.53e11 (which is massive).

I tried it and I get the following error:

Error using zeros
Requested array exceeds the maximum possible variable size.

How can you deal with such a scenario?

Answer 2

[Tom Lane]

It seems like 1:46 represents your possible data values and [n1 n2 ... n46] represents the frequency of each value. If you "edit mle" and look toward the bottom, it has a little function that expands this sort of representation into a vector with n1 copies of 1, n2 copies of 2, and so on. You could try copying this code to your own function. A warning is that you will need to remove values with a frequency of zero before calling the function.

If you're not comfortable doing this, or if I haven't answered your question, let me know.

[New]

Couldn't make it work properly...

Some easier way?

Thanks