Statalist


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

RE: st: keeping likelihood for every observation


From   "Nick Cox" <[email protected]>
To   <[email protected]>
Subject   RE: st: keeping likelihood for every observation
Date   Sun, 7 Feb 2010 16:50:59 -0000

Here is another way to do it. 

Before your likelihood code, initialise a variable to save the likelihood 

gen ml_save = . 

Within your likelihood code, follow each update of `lnf' by 

replace ml_save = `lnf' 

After running your likelihood code, investigate the saved variable. 

This pivots on the simple idea that a permanent variable remains in memory until -drop-ped explicity. Naturally, use a name that does not clash with others, and beware that use of e.g. -preserve- might frustrate this. This will add a bit to execution time. 

Nick 
[email protected] 

Maarten buis

--- On Sun, 7/2/10, Stephane Mahuteau wrote:
> Could anyone let me know how I can save the log likelihood
> for every observation in my dataset and keep it as a
> variable? 

You can recreate the likelihood relatively simply with the
-predictnl- command, as in the example below. Whether that
is efficient depends on how you want to use that variable.

*--------------------- begin example -----------------------------------
program drop _all
program define mynorm_lf
        args lnf xb ln_sigma
        qui replace `lnf' = lnnormalden($ML_y1, `xb', exp(`ln_sigma'))
end
sysuse auto, clear
ml model lf mynorm_lf (mu: mpg = foreign rep78 displacement) /ln_sigma
ml check
ml search
ml maximize

predictnl double lnf = lnnormalden(mpg,xb(#1),exp(xb(#2))) if e(sample)

// check if this is the correct variable
sum lnf, meanonly
di r(sum)
*----------------------- end example -----------------------------------
( For more on how to use examples I sent to statalist see:
 http://www.maartenbuis.nl/stata/exampleFAQ.html )

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/



© Copyright 1996–2024 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index