Bookmark and Share

Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at

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

Re: st: ml mixture distribution

From   Maarten Buis <>
Subject   Re: st: ml mixture distribution
Date   Tue, 10 Jul 2012 09:55:56 +0200

On Mon, Jul 9, 2012 at 6:39 PM, Lukas Kornher wrote:
>I know the command. I have a mixture distribution of a
> normal and two truncated normal which is not available for fmm. Do you think
> it is possible to rewrite Deb's original code?

Sure, I have learned most of my Stata programming from looking at
other person's code for similar problems. You can see the code by
typing -viewsource fmm.ado-.

Another useful strategy is to first program a simpler problem. I would
start with creating the programs inside a .do file and begin with just
a normal distribution without any mixture components, check that
against -regress-, than extend that to a problem that is already
implemented in -fmm-, e.g. a normal distribution with two normal
components. See if that works and gives the same results as -fmm-,
than complicate it by changing the components to truncated normals,
and than add bells and whistles like turning this into an .ado file
and checking for valid input.

Below is an example of how I would begin such a project.

*---------------------- begin example ------------------------
// clear programs and data
clear all

// create log-likelihood evaluator
program define mynormal_lf
	args lnf mu ln_sigma
	tempname sigma
	scalar `sigma' = exp(`ln_sigma')
	qui replace `lnf' = ln(normalden($ML_y1, `mu', `sigma'))

// open example data
sysuse auto

sum weight if foreign == 0, meanonly
gen c_weight = (weight - r(min))/2000
label var c_weight ///
    "weight centered at lightest domestic car (short tons)"

gen c_price = (price - 6000) / 1000
label var c_price "price centered at 6000$ (1000$)"

// specify the model
ml model lf mynormal_lf ///
   mu: c_price = c_weight foreign) /ln_sigma

// check and maximize
ml check
ml search
ml maximize

// check against regress
regress c_price c_weight foreign
*----------------------- end example -------------------------
(For more on examples I sent to the Statalist see: )

Apart from that, you will need excess to:
William Gould, Jeffrey Pitblado and Brian Poi (2010) "Maximum
Likelihood Estimation with Stata", fourth edition. College Station,
TX: Stata Press.

Hope this helps,

Maarten L. Buis
Institut fuer Soziologie
Universitaet Tuebingen
Wilhelmstrasse 36
72074 Tuebingen
*   For searches and help try:

© Copyright 1996–2017 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   Site index