GraphPad Statistics Guide

How to: Monte Carlo analyses

How to: Monte Carlo analyses

Previous topic Next topic No expanding text in this topic  

How to: Monte Carlo analyses

Previous topic Next topic JavaScript is required for expanding text JavaScript is required for the print function Mail us feedback on this topic!  

How to begin a Monte Carlo analysis

Simulate a data table using one of Prism's simulation analyses.. Note that these simulations include random scatter, so will produce new results when they are updated.

1.Analyze that simulated data set as appropriate.

2.From that results page, click Analyze and choose Monte-Carlo analysis. This analysis will repeat the simulations many times, and tabulate selected results. The Monte Carlo analysis will only be available for analyses that create analysis constants. Note that linear regression does not, but you can fit a straight line with the nonlinear regression analysis.

The explanations below explain the basic ideas of the Monte Carlo analysis. Follow the exampleto learn the details.

Simulations tab

How many simulations?

How many simulations should you run?  If you make only a few simulations, the results will be affected too much by chance. Running more simulations will give you more precise results, but you'll have to wait longer for the calculations to repeat. When just playing around, it might make sense to use as few as 100 simulations so you can see the results instantly. When trying to polish simulation results, it can make sense to use as many as 10,000 or 100,000 simulations. A good compromise is 1000.

Append?

If you go back to run more simulations, check an option box to append the new simulations to the existing results, rather than start fresh.

Random seed

The choice of random numbers used in a series of simulations depends on the random number seed used to generate the first set of results. By default, Prism picks this seed automatically (based on the time of day), and presents this seed in a floating note superimposed on Monte Carlo results.

If you want two or more Monte Carlo analyses to use precisely the same data sets (so you can compare two ways of analyzing those data), enter that random seed on the Simulation tab.

Parameters to tabulate tab

Prism lists all of the analysis constants generated by the analysis. Check the ones whose values you want to tabulate.

You cannot change the set of values included in this list. Let us know if there are parameters missing, and we can add them in a future version.

Hits tab (optional)

If you skip this Hits tab, Prism will tabulate the selected parameters (different columns) for each simulation (rows).

Prism can also reduce the results down to a single number -- the fraction of the simulations that are "hits". Define a hit to be when a value tabulated by the analysis equals a certain value or is within a specified range. Click New...Graph of existing data from this table, and choose a parts-of-whole graph to create a pie graph of the fraction of hits vs. not hits.

Prism can also tabulate the selected parameters only for simulations that are hits, and/or for only the simulations that are not hits. Choose any or all of these options (Hits, Not hits, All simulations) at the bottom of the Hits tab. Each option you check will create its own results table.