In an unpublished note, Sundar Dorai-Raj recently cited the following as addressing this question:

Cox, D. R. and Snell, E. J. (1989) The Analysis of Binary Data, Second Edition, London: Chapman and Hall.

Nagelkerke, N. J. D. (1991) “A Note on a General Definition of the Coefficient of Determination,” Biometrika, 78, 691 -692.

Apparently, Cox and Snell (1989) suggest the following

R1.2 = 1-(L(0)/L(b.hat))^(2/n),

where L(b) = log(likelihood(b)). With a normal likelihood using the standard maximum likelihood estimate for the variance, this produces the standard formula for the coefficient of determination.

Nagelkerke (1991) suggested the following modification:

R2.2 = R1.2/(1-(L(0))^(2/n))

I don't understand this second formula, so I can't comment on it. Yesterday, I found a few more recent papers that looked potentially relevant in a search of "query.statlib.org" for "coefficient of determination". However, I won't know if they are relevant until I actually see them.

Hope this helps.
Best Wishes,
Spencer Graves

Allin Cottrell wrote:
On Mon, 17 Mar 2003, Daniel Bloch wrote:


I analysed data with LME in R. Is there a measure for LME
(likelihood estimated) statistics which has an analogous meaning to
the coefficient of determination (r-square) estimated by
least-square procedure?


There is not an exact analog, but the log-likelihood is commonly used
as a figure of merit.

Allin Cottrell.

______________________________________________
[EMAIL PROTECTED] mailing list
https://www.stat.math.ethz.ch/mailman/listinfo/r-help

______________________________________________ [EMAIL PROTECTED] mailing list https://www.stat.math.ethz.ch/mailman/listinfo/r-help

Reply via email to