Monte Carlo simulation from a triangular distribution in Python

Question

I'm new to python and trying to use the numpy.random triangular function to run a series of Monte Carlo simulations from several triangular distributions and then append the simulation outputs from each run. The sample data is as below.

ID  Low  Mode  High  
A   10   15    25
B   7    20    22
C   2    18    20
D   1    4     5
E   13   25    34

I would like to run 10000 runs for each ID and append the results. I know I can run for each ID for example ID A using np.random.triangular(10, 15, 25, 10000). May need to write for loop to run and append all IDs. Thank you!

Update!

The expected output format is:

ID Run      Output
A   1       11
A   2       23
.
.
.
A   10000   18
B   1       21.5
B   2       9
.   .       .
.   .       .
.   .       .
B   10000   19
C   1       2.5
C   2       13
.   .       .
.   .       .
.   .       .

Could you give an example of the output you want so I know how to write the answer? — DPM, Oct 28 '20 at 18:40
@DPM - I updated my post with the expected output. Thank you! — BMM, Oct 28 '20 at 18:51
Isn't it simpler if there is a list for every ID? Also are you using the library pandas? — DPM, Oct 28 '20 at 18:59
@DPM - I prefer one dataframe with all the output so that I can summarize (like mean) for each ID in one table. I have more than 50 IDs in my dataset. — BMM, Oct 28 '20 at 19:05
The run column doens't make mauch sense because the numpy function gives you a sample, it only runs once — DPM, Oct 28 '20 at 19:17
I have written the code you want but it takes a long time to run, think carefully if you really need the run column because it is unnecessary — DPM, Oct 28 '20 at 20:01
@DPM - the Run column is not a must as long as I get the other two columns (ID and Output) as an output. Thank you! — BMM, Oct 28 '20 at 20:08
Ok, I'll just check if the code is write and then upload the answer — DPM, Oct 28 '20 at 20:09
One thing I suggest is instead of a row per run have a row per index and in each row the output will be an array of the sample — DPM, Oct 28 '20 at 20:12
@DPM - Thank you! I tried your code and got the following error message raise TypeError("data argument can't be an iterator") TypeError: data argument can't be an iterator ... the error occurs when you assign df. — BMM, Oct 28 '20 at 20:34
It's because of zip. Check this https://stackoverflow.com/questions/45388800/python-data-argument-cant-be-an-iterator/45388890 — DPM, Oct 28 '20 at 21:14
@DPM - thanks! It works now. I modified it to get what I wanted. Thanks again! — BMM, Oct 29 '20 at 18:03

DPM · Accepted Answer · 2020-10-28T20:18:30.493

df is a dataframe with your data and in the cycle I am iterating through the rows, so basically each row of the new dataframe will have an array with the 10000 samples.

import pandas as pd
import numpy as np
Low = [10,7,2,1,13]
Mode = [15,20,18,4,25]
High = [25,22,20,5,34]
ID = ['A', 'B', 'C', 'D', 'E']
df = pd.DataFrame(zip(Low, Mode, High), columns = ['Low', 'Mode', 'High'], index = ID)
cols = ['Output']
        
df2 = pd.DataFrame(columns=cols, index = ID)

for l in range(5):
    result = np.random.triangular(df.iloc[l][0], df.iloc[l][1], df.iloc[l][2], 10000)
    df2.iloc[l][0] = result

An example of the output:

Monte Carlo simulation from a triangular distribution in Python

1 Answers1