Interpreting StatsChi Test, specifically p-value

Supriya Balaji wrote:

statschitest/S/T=1 observed hist ,expected hist

This command would not take you very far. I assume you meant

statschitest/S/T=1 observedHist ,expectedHist

or some variation thereof.

To understand the P-value I suggest that you try the following experiment:
1. Create a simple distribution:
Make/O/N=200 ggg
SetScale/P x -100,1,"", ggg
ggg=100*exp(-x^2/400)
Display ggg

2. Create a test distribution by copying ggg and adding some noise:

Duplicate/O ggg,exg
exg=round(exg+enoise(1))
appendTograph exg

3. Run the test:

StatsChiTest/T=1/S exg,ggg

4. Now increase the noise and test again:

exg=abs(round(exg+enoise(2)))
StatsChiTest/T=1/S exg,ggg

You can draw your own conclusions about P as a function of the noise level.

• Log in or register to post comments