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

Re: AW: st: 3 Level - LMM - Number of clusters?

From   Steven Samuels <>
Subject   Re: AW: st: 3 Level - LMM - Number of clusters?
Date   Sun, 29 Jun 2008 13:54:18 -0400


Do simpler models work?

1. Ignore SIC and just do a nesting within country?
2. Ignore country and just nest in coarse and refined SIC groupings, coded to not repeat between countries).
3. Fit a fixed effect for country and a random effect for SIC, also coded not to repeat between counries?

What commands are you using? What models and covariates?


On Jun 25, 2008, at 5:21 AM, Christian Weiß wrote:

Dear Steven,

It's almost as you described in "B".

Firms are nested within in industries
& Industries are nested within countries.

You are definitely right that SIC codes are not country specific. However, the data I am analyzing do take country differences of industries into account, so there is no problem at this point.

At this point, the key question is whether there might be a problem with the number of clusters I am trying model. (2000 firms, 11 countries, 9 (or 69) industries).

Best regards
-----Ursprüngliche Nachricht-----
Von: [mailto:owner-] Im Auftrag von Steven Samuels
Gesendet: Mittwoch, 25. Juni 2008 03:41
Betreff: Re: st: 3 Level - LMM - Number of clusters?

Do you mean:
1. countries are nested within industries
2. firms are nested within countries?

This doesn't seem right to m
1. industries are nested within countries
2. firms are nested within industries.

In fact, neither A nor B seems right to me, as SIC codes are not
unique to countries nor countries to SIC codes.

Please clarify the structure of your problem.

On Jun 24, 2008, at 12:04 PM, Christian Weiß wrote:


I am currently trying to bugfix a linear mix model. My sample
comprises about 2000 observations (firms) which are nested as follows:

Level 1: nested in 9 industries (1 Digit SIC Code) or(!) 68
industries (2 digit SIC code)
Level 2: nested in 11 countries

Most of the modeled effects will be random.

Is it possible, that the number of industries / countries causes
my model not to converge? (I somehow recall having read that about
20 clusters should be used?)

*   For searches and help try:

*   For searches and help try:

*   For searches and help try:

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