[Top][All Lists]

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

Re: PSPP-BUG: pspp bug

From: John Darrington
Subject: Re: PSPP-BUG: pspp bug
Date: Fri, 6 Apr 2012 08:26:53 +0000
User-agent: Mutt/1.5.18 (2008-05-17)

Now that you point it out, I see that I'm getting different failures on
different machines.

I think the root cause is a near singular matrix of this example, which
is causing the determinant to be very close to zero.  Note that log 
isn't all that  large (-36.043653)

So perhaps we should just clamp values below DBL_EPSILON to zero.

Also, I see that there are a lot of memory leaks in FACTOR which we need to fix.


On Thu, Apr 05, 2012 at 10:12:53PM -0700, Ben Pfaff wrote:
     When we looked at this in IRC, I think we were looking at the
     wrong value.  You were talking about log(0) but that's involved
     in "Bartlett's Test of Sphericity" not in "Kaiser-Meyer-Olkin
     Measure of Sampling Adequacy".
     I added
           printf ("sum_ssq_r=%e sum_ssq_a=%e\n", sum_ssq_r, sum_ssq_a);
     as line 1870 of factor.c and got:
             sum_ssq_r=3.472064e+35 sum_ssq_a=1.036522e-30
     so that the calculation sum_ssq_r /  (sum_ssq_r + sum_ssq_a) is
     basically x / (x + 0) == 1, which seems reasonable.  What do you
     see when you add that line?

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.

Attachment: signature.asc
Description: Digital signature

reply via email to

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