Generate values from a frequency distribution

Question

I'm currently analyzing a 16 bit binary string - something like 0010001010110100. I have approximately 30 of these strings. I have written a simple program in Matlab that counts the numbers of 1's in each bit for all 30 strings.

So, for example:

1 30

2 15

3 1

4 10

etc

I want to generate more strings (100s) that roughly follow the frequency distribution above. Is there a Matlab (or Python or R) command that does that?

What I'm looking for is something like this: http://www.prenhall.com/weiss_dswin/html/simulate.htm

If you only have 30 strings how does 2 (I assume that's a bit position) get a count of 45? — PM 2Ring, Apr 28 '17 at 08:09
I don't get the logic that you are trying to explain. Why not just show your real data, and the output that you expect? — Bernhard, Apr 28 '17 at 08:21
Thanks for fixing that confusing sample data. In Python 3.6+ you can use [`random.choices`](https://docs.python.org/3/library/random.html#random.choices). — PM 2Ring, Apr 28 '17 at 08:23

score 0 · Accepted Answer · edited Apr 28 '17 at 08:44

In MATLAB: just use < (or lt, less than) on rand:

len = 16; % string length
% counts of 1s for each bit (just random integer here)
counts = randi([0 30],[1 len]); 
% probability for 1 in each bit
prob = counts./30;
% generate 100 random strings 
n = 100;
moreStrings = rand(100,len);
% for each bit check if number is less than the probability of the bit
moreStrings = bsxfun(@lt, moreStrings, prob); % lt(x,y) := x < y

In Python:

import numpy as np

len = 16 # string length
# counts of 1's for each bit (just random integer here)
counts = np.random.randint(0, 30, (1,16)).astype(float)
# probability for 1 in each bit
prob = counts/30
# generate 100 random strings 
n = 100
moreStrings = np.random.rand(100,len)
# for each bit check if number is less than the probability of the bit
moreStrings = moreStrings < prob

Generate values from a frequency distribution

1 Answers1