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]

st: Exceeding maxvar in nested anova estimation

From   "Wissoker, Doug" <>
To   <>
Subject   st: Exceeding maxvar in nested anova estimation
Date   Thu, 3 Feb 2011 14:14:24 -0500

Hi all,

I'm trying to estimate a nested anova model -- modeling variation (in
program enrollment) by state (statefip) and Public Use Microdata Areas
(puma_index) within state.  Altogether there are 51 state groups and 2069
PUMAs.  (Each PUMA code is found within only one state.)

I'm using the command:

anova r1 statefip / puma_index|statefip , dropemptycells

When I use the full sample, I get the error:  "maxvar too small    You have
attempted to use an interaction with too many levels or attempted to fit a
model with too many  variables.  You need to increase maxvar; it is
currently 32767.  Use set maxvar; see help maxvar."

Page 46 of the base manual explains that with the nested model, you can get
a design matrix that is too large and suggests that it can be solved using
the dropemptycells option.   ("We suggest that you specify dropemptycells
when there are nested terms in your ANOVA.")    But I used the
dropemptycells option and still have the problem.

I was only able to estimate the model when I dropped PUMAs to the point
where the product of the numbers of PUMAs and states is less than 32767 or

Any suggestions would be welcome.  (I'd like to use the full set of
areas.)   Please let me know if additional information is needed to answer
my question.   (I'm using Stata 11.1 for windows 64 bit x86-64 with a

Thanks much,
Doug Wissoker

*   For searches and help try:

© Copyright 1996–2018 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   Site index