Iman-Conover and matrices of 1s
From: J. Mitchell (donotspam_at_me.plz)
Date: 03/15/05
- Next message: htothefro_at_gmail.com: "Re: estimating parameters in a Box-Jenkin's transfer function"
- Previous message: dave_at_autobox.com: "Re: estimating parameters in a Box-Jenkin's transfer function"
- Next in thread: glenbarnett_at_geocities.com: "Re: Iman-Conover and matrices of 1s"
- Reply: glenbarnett_at_geocities.com: "Re: Iman-Conover and matrices of 1s"
- Messages sorted by: [ date ] [ thread ]
Date: Tue, 15 Mar 2005 20:24:31 +0200
I am using the Iman-Conover method of inducing rank correlation. However,
due to it using a Choleski decomposition, the following matrix will fail to
due zero being divided by zero, Lij = (Aji - Sum) / Lii:
A B C
A 1 1 1
B 1 1 1
C 1 1 1
If I have the Choleski algorithm interpret 0/0 (which is not defined) as 0,
I [seem] to get correct results, however is this mathematically correct? If
not, what can I do about it.
Also, how can I modify an inconsistent matrix (the one given above *is*
consistent) to be consistent?
TIA
- Next message: htothefro_at_gmail.com: "Re: estimating parameters in a Box-Jenkin's transfer function"
- Previous message: dave_at_autobox.com: "Re: estimating parameters in a Box-Jenkin's transfer function"
- Next in thread: glenbarnett_at_geocities.com: "Re: Iman-Conover and matrices of 1s"
- Reply: glenbarnett_at_geocities.com: "Re: Iman-Conover and matrices of 1s"
- Messages sorted by: [ date ] [ thread ]
Relevant Pages
|
|
