Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: st: xtmixed with nonrtolerance. What happens?
From
Stas Kolenikov <[email protected]>
To
[email protected]
Subject
Re: st: xtmixed with nonrtolerance. What happens?
Date
Wed, 22 Jun 2011 18:48:22 -0400
It looks like you have data with wildly differing scales. I understand
that you need to interpret the results in the original scales, but
maybe you could rescale your variables so that all of your
coefficients would be about 1. Whether that will help convergence is
anybody's telling, of course, but usually differences in the scales
(and hence coefficients) of the order of 1e3-1e4 are detrimental to
numeric convergence.
On Wed, Jun 22, 2011 at 4:33 PM, "Lukas Bösch" <[email protected]> wrote:
> Dear Statalist community.
>
> I am using Stata 10.0 and doing a mixed model analysis of export data.
> After trying different options and always having trouble to get a propper output i finally found a way to get to my results. I however could not find any information about why it works and if it is allright. But let us first start with the problem:
>
> 1) This is the command i enter and the output stata creates:
>
> xtmixed quantity year centforestarea2 centgdp2 centlandarea2 centpopulation2 || _all: R.country || _all: R.genus
>
> Performing EM optimization:
>
> Performing gradient-based optimization:
>
> Iteration 0: log restricted-likelihood = -77051.164
> Iteration 1: log restricted-likelihood = -77046.704
> Iteration 2: log restricted-likelihood = -77046.565
> Iteration 3: log restricted-likelihood = -77046.5
> Iteration 4: log restricted-likelihood = -77046.468 (backed up)
> Iteration 5: log restricted-likelihood = -77046.46 (backed up)
> Iteration 6: log restricted-likelihood = -77046.456 (backed up)
> Iteration 7: log restricted-likelihood = -77046.454 (backed up)
> numerical derivatives are approximate
> nearby values are missing
> Iteration 8: log restricted-likelihood = -77046.453 (backed up)
> numerical derivatives are approximate
> nearby values are missing
> Hessian has become unstable or asymmetric
>
> Mixed-effects REML regression Number of obs = 6192
> Group variable: _all Number of groups = 1
>
> Obs per group: min = 6192
> avg = 6192.0
> max = 6192
> Wald chi2(5) = 9.26
> Log restricted-likelihood = -77051.164 Prob > chi2 = 0.0991
> quantity | Coef. Std. Err. z P>|z| [95% Conf. Interval]
> year | -429.7599 215.8898 -1.99 0.047 -852.8961 -6.623654
> centfores~a2 | -9875.264 6631.861 -1.49 0.136 -22873.47 3122.945
> centgdp2 | -2024.629 4138.469 -0.49 0.625 -10135.88 6086.621
> centlandar~2 | -52889.76 63817.96 -0.83 0.407 -177970.7 72191.13
> centpopul~n2 | 22296.98 10234.72 2.18 0.029 2237.304 42356.66
> _cons | 895402.2 433369.4 2.07 0.039 46013.74 1744791
>
> Random-effects Parameters | Estimate Std. Err. [95% Conf. Interval]
>
> _all: Identity |
> sd(R.country) | 313329.2 .
> _all: Identity |
> sd(R.genus) | 6757.304 .
> sd(Residual) | 60169.26 .
> LR test vs. linear regression: chi2(2) = 7810.42 Prob > chi2 = 0.0000
>
> Note: LR test is conservative and provided only for reference.
> Warning: convergence not achieved; estimates are based on iterated EM
>
> Obviously Stata has a problem and can't calculate the standard errors of the random factors.
>
> 2) With the option nonrtolerance it works however:
>
> xtmixed quantity year centforestarea2 centgdp2 centlandarea2 centpopulation2 || _all: R.country || _all: R.genus, nonrtolerance
>
> Performing EM optimization:
>
> Performing gradient-based optimization:
>
> Iteration 0: log restricted-likelihood = -77051.164
> Iteration 1: log restricted-likelihood = -77046.704
> Iteration 2: log restricted-likelihood = -77046.565
> Iteration 3: log restricted-likelihood = -77046.5
> Iteration 4: log restricted-likelihood = -77046.468 (backed up)
> Iteration 5: log restricted-likelihood = -77046.46 (backed up)
> Iteration 6: log restricted-likelihood = -77046.456 (backed up)
>
> Computing standard errors:
>
> Mixed-effects REML regression Number of obs = 6192
> Group variable: _all Number of groups = 1
>
> Obs per group: min = 6192
> avg = 6192.0
> max = 6192
>
>
> Wald chi2(5) = 9.22
> Log restricted-likelihood = -77046.456 Prob > chi2 = 0.1008
> quantity | Coef. Std. Err. z P>|z| [95% Conf. Interval]
> year | -429.7645 216.4073 -1.99 0.047 -853.915 -5.614053
> centfores~a2 | -9885.307 6647.52 -1.49 0.137 -22914.21 3143.592
> centgdp2 | -2021.312 4148.464 -0.49 0.626 -10152.15 6109.527
> centlandar~2 | -52859.75 63778.66 -0.83 0.407 -177863.6 72144.12
> centpopul~n2 | 22276.96 10257.46 2.17 0.030 2172.715 42381.2
> _cons | 895338.1 434389.3 2.06 0.039 43950.68 1746726
>
> Random-effects Parameters | Estimate Std. Err. [95% Conf. Interval]
> _all: Identity |
> sd(R.country) | 313133.2 36075.6 249840.9 392459.4
> _all: Identity |
> sd(R.genus) | 3440.288 1355.694 1589.157 7447.712
> sd(Residual) | 60315.87 545.9681 59255.23 61395.5
> LR test vs. linear regression: chi2(2) = 7819.83 Prob > chi2 = 0.0000
> Note: LR test is conservative and provided only for reference.
>
> Can someone explain to me why it works with nonrtolerance and tell me if these outputs are as reliable as if they were created without nonrtolerance. I searched in the stata help and on stata.com but could not find more information about this.
>
> Kind regards
>
> Lukas
>
> --
> NEU: FreePhone - kostenlos mobil telefonieren!
> Jetzt informieren: http://www.gmx.net/de/go/freephone
> *
> * 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/
>
--
Stas Kolenikov, also found at http://stas.kolenikov.name
Small print: I use this email account for mailing lists only.
*
* 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/