One-Way ANOVA in Inference App
12-19-2016, 10:10 PM
Post: #1
 mark4flies Member Posts: 115 Joined: Jan 2015
One-Way ANOVA in Inference App
I am going through the User Guide to test the latest release and for my own review. I entered some data for four groups like this into the Inference app:

MAKELIST(RANDNORM(50,3),X,1,10) STO> I1 (through I3) and
MAKELIST(RANDNORM(53,3),X,1,10) STO> I4

So I have generated random samples of size 10 from four populations, where the last of them is different by a mean of 3. Then I tested the null hypothesis with the one-way ANOVA. I confirmed this result in the Num view:

F: 3.207 (correct)
P: 0.03446 (correct)

But I was surprised to see this result in the Plot view:
ChiSquare: -10.8849 (?)
P: 0.03446 (correct)

My questions are:
Why is Chi square in the plot view?
Where does the Chi square statistic come form?
Is the graph in the Plot view for a Chi square distribution or the F distribution?

Thanks!
12-20-2016, 12:49 AM
Post: #2
 Tim Wessman Senior Member Posts: 2,293 Joined: Dec 2013
RE: One-Way ANOVA in Inference App
I'm guessing it might be plotting when it shouldn't there. I'll take a look at it. I don't think there is a plot view currently for Anova. Thanks!

TW

Although I work for HP, the views and opinions I post here are my own.
12-21-2016, 01:55 PM
Post: #3
 mark4flies Member Posts: 115 Joined: Jan 2015
RE: One-Way ANOVA in Inference App
Well, a plot view with the F PDF and the sample statistic, as done for t tests, might be great for interpretation and instruction of ANOVA.
12-21-2016, 02:40 PM (This post was last modified: 12-21-2016 02:41 PM by Tim Wessman.)
Post: #4
 Tim Wessman Senior Member Posts: 2,293 Joined: Dec 2013
RE: One-Way ANOVA in Inference App
I remember implementing this, so it was strange it wasn't displaying right.

I looked into the source today. Looks like the root culprit was a missing "break" in a switch statement which causes the *label* to display incorrectly as chi2. The graph itself is correct, but the wrong text for the front half/value are being shown. Testers missed this one (and I screwed it up of course ). Thanks!

It should read DF:<num> F:<val> P:<val>.

TW

Although I work for HP, the views and opinions I post here are my own.
 « Next Oldest | Next Newest »

User(s) browsing this thread: 1 Guest(s)