[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Correlation

From: Ben Pfaff
Subject: Re: Correlation
Date: Sat, 25 Oct 2008 12:44:06 -0700
User-agent: Gnus/5.11 (Gnus v5.11) Emacs/22.2 (gnu/linux)

"Richard MacLennan" <address@hidden> writes:

>     I read some of your online blogs, but I could not find a simple way
> to contribute to the discussion, so I thought I would do so through email.

I didn't know that any of us had blogs.  Email is the best choice

> I am using version 0.4.0, whcih is what you seem to be
> distributing with PSPPIRE.

There is some confusion here.  The current version of PSPP is
0.6.1.  Only PSPP versions 0.6.0 and 0.6.1 include the PSPP GUI,
named PSPPIRE.  So, if you have a GUI, then you have PSPP 0.6.0
or later.

> 1. Someone requested correlations. Someone else recommend using the
> correlation option in the CROSSTAB procedure in the Descriptives menu. I
> would advise against this. The default correlations it reports (phi &
> Cramer's V) are only relevant for 2 x 2 and larger contingency tables,
> respectively. This would not be suitable for the most common
> applications of correlation.
> 2. To obtain correlations, instead I would recommend using the PAIRED
> TTEST procedure in the Compare Means menu. The correlation reported here
> is the equivalent of a PEARSON CORR between the two paired variables.
> 3. BTW, I would not trust the significance test for the t-test value in
> the PAIRED procedure. I tested in with the Employee.sav file from SPSS,
> and it reports p = 2.000. I hope that even non-statisticians would
> realize this value is incorrect.

Thank you for these suggestions.  I am not sure how to act on
them.  Do you think that these suggestions are suitable for
adding to the manual, as documentation, or do you have another

> 4. In your documentation you do not really clarify whether PEARSON CORR
> is an implemented or non-implemented procedure. It appears to be
> non-implemented. However, it would not seem hard to do so, since you
> already have it coded for the PAIRED TTEST procedure (although of course
> you do not have the complete syntax for PEARSON CORR coded).

For what it's worth, PEARSON CORR and CORRELATIONS are two names
for the same procedure, but it is not implemented under either
name.  Based on your logic, it sounds like we should consider
implementing it soon.
Ben Pfaff

reply via email to

[Prev in Thread] Current Thread [Next in Thread]