CODEWITHSUNDEEP
matlabstatisticshistogramprobabilitycdf
Sari
Sari

Reputation: 101

How to make a CDF from a histogram in MATLAB

I am trying to create function that takes a histogram and makes a CDF from it. However I cannot use the cdfplot function in Matlab.

How would I go about doing this?

This produces the input histogram:

x = randn(1000,1);
nbins = 25;
h = histogram(x,nbins)

Upvotes: 4

Views: 13012

Answers (2)

ibezito
ibezito

Reputation: 5822

Use MATLAB's cumsum function, after normalizing the histogram.

hNormalized = h.Values / sum(h.Values);
cdf = cumsum(hNormalized)

The full code:

x = randn(1000,1);
nbins = 25;
h = histogram(x,nbins);
hNormalized = h.Values / sum(h.Values);
cdf = cumsum(hNormalized);

Results with smaller nBins (nBins = 8):

hNormalized =

0.0210    0.0770    0.1930    0.2830    0.2580    0.1250    0.0370    0.0060

cdf =

0.0210    0.0980    0.2910    0.5740    0.8320    0.9570    0.9940    1.0000

Upvotes: 2

horchler
horchler

Reputation: 18484

The most straightforward way to create a cumulative distribution from data is to generate an empirical CDF. The ecdf can do this directly. By default, this doesn't require one to produce a histogram for a dataset:

x = randn(1000,1);
ecdf(x);

Empirical CDF

However, if you want a lower resolution CDF, you can use histogram directly with the 'cdf' normalization option:

x = randn(1000,1);
nbins = 25;
histogram(x,nbins,'Normalization','cdf');

CDF histogram option

You might find the 'cumcount' option useful too. See the documentation for ecdf and histogram for details on how to extract and use the output from these functions.

Upvotes: 3

Related Questions