Re: PSPP-BUG: BUG in FACTOR

[Douglas Bonett]

Here is a screen output of the results. 

On Tue, Sep 10, 2013 at 11:09 PM, John Darrington <address@hidden> wrote:

On Tue, Sep 10, 2013 at 08:44:03PM -0700, Douglas Bonett wrote:
     With principal axis factoring of a correlation matrix with the following
     9x3 data matrix

         1.00    5.00    3.00
         2.00    2.00    2.00
         3.00    1.00    1.00
         4.00    4.00    5.00
         5.00    3.00    9.00
         6.00    6.00    4.00
         7.00    7.00    6.00
         8.00    8.00    8.00
         9.00    9.00    7.00

     PSPP 0.8 gives loading estimates of 0 0 0. They should be .990  .808  .723.

When I try it, I get exactly the results you say it should be.  This is what I tried,
does it not work for you?

set format = F20.3.
set decimal=dot.
data list list /x y z *.
begin data.

1.00    5.00    3.00
    2.00    2.00    2.00
    3.00    1.00    1.00
    4.00    4.00    5.00
    5.00    3.00    9.00
    6.00    6.00    4.00
    7.00    7.00    6.00
    8.00    8.00    8.00
    9.00    9.00    7.00

end data.

factor /variables = x y z
        /method = correlation
        /extraction = paf.

     Also, when returning to the GUI setup menu, the previous specification is
     lost and must be re-entered.

Yes.  This is on the list of "wishlist bugs". See http://savannah.gnu.org/bugs/?31588
If anyone wants to fix this and submit a patch, I'm sure Ben will happily accept it.

J'



--
PGP Public key ID: 1024D/2DE827B3
fingerprint = 8797 A26D 0854 2EAB 0285  A290 8A67 719C 2DE8 27B3
See http://keys.gnupg.net or any PGP keyserver for public key.

FAoutput.pdf
Description: Adobe PDF document