MathBench > Statistical Tests

Goodness of Fit Tests

Let's ask the computer to do more...

For a person clicking on a button and writing down results, 10 or 20 trials is a lot. For a computer to run 10 or 20 trials and keep track of the results requires about a millisecond of processor time.

So let's ask the computer to do more. 100 trials? 1000 trials? Sure, no problem.

In the applet below, when you click the "Simulate 1 year" button, the computer will run 1 trial of 100 sickdays (i.e., one sick-year), count how many fell on Monday or Friday, and put a bar on the chart in the appropriate spot.



Number of Monday/Fridays: --
Help me with graph
empty graph

 

First, simulate a "year" worth of sickdays. How many M/F sickdays were there?

Keep going a year at a time until the "100 years" button lights up.

Click the "100 years" button. What is the average # of M/F sickdays?

Do the results in the applet support or reject Dilbert's hypothesis (that sickdays are random)?

 

(To make this problem interactive, turn on javascript!)

I think I have the answer: Most of the time when you run this applet,
AT LEAST a third of the trials have 42 or more mon/fri sickdays,
which makes it very common -- this supports the null hypothesis,
which was that the data fit a random model.