[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

Re: st: weird error in -nlcom-

From   [email protected] (Roberto G. Gutierrez, StataCorp)
To   [email protected]
Subject   Re: st: weird error in -nlcom-
Date   Mon, 07 Jan 2008 11:34:44 -0600

Alan H. Feiveson <[email protected]> writes:

> Hi - In Stata 10 (Windows), after runnning a -ml- program (of my own), I
> tried to use -nlcom- .

> Depending on what values I give it (see below) it either works, or gives a
> peculiar error messgae " maximum number of iterations exceeded".

> First of all, -nlcom- is supposed to use the delta method, which doesn't
> involve iterating anything - so what's going on?

> Even more mystifying is that if I slightly change the inputs, it works or
> doesn't work (see below)

-nlcom- does utilize the delta method, but the derivatives therein are
numerical.  Calculating a numerical derivate requires you select a step size,
and the process of choosing an optimal one is iterative.  Choose one too small
and you are (practically) dividing by zero; choose one too big and you are
smoothing over important local features of the function in question.

It is somewhat rare that the iterative process of choosing the optimal step
size will exceed the maximum iterations, but it does happen on occasion.  In
these cases, one can usually sniff out the culprit, such as trying a estimate
a derivative that is not exactly zero (easy to detect) but very very close to

If Alan wishes to email me privately with the details and data, I'll be glad
to take a closer look.

[email protected]
*   For searches and help try:

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