Page 1 of 1
Getting p-values and z-ratios after MCMC
Posted: Tue May 22, 2012 10:49 am
by AnjaScheiwe
Hi,
I am running cross-classified models using the MCMC option. The output shows the z values and ESS. The runmlwin help file says it is possible to get p-values with the option zratio, but when I try I get this error message:
'option zratio not allowed'
This is how I specified the option:
Code: Select all
runmlwin read cons , level3 (dlsoa:cons) level2(schoolid: cons) level1(mcsid: cons) mcmc (cc ) initsb (b) zratio nopause
And I noticed that the column header for what I think must be standard errors says "Std. Dev."?
Many thanks for your help!
Anja
Re: Getting p-values and z-ratios after MCMC
Posted: Tue May 22, 2012 12:10 pm
by GeorgeLeckie
Hi Anja,
This works for me (see below). Perhaps you do not have the latest version of runmlwin? You can make sure you have the latest version by reinstalling the command
. ssc install runmlwin, replace
You have fitted the model by MCMC. The reported parameter estimate and standard error are the mean and standard deviation of each parmater chain. This is why the column header says Std. Dev. rather than Std. Err. For more details, see ...
The following cross-classified MCMC model where I specify the zratio option works for me...
Code: Select all
use http://www.bristol.ac.uk/cmm/media/runmlwin/xc, clear
rename sex female
matrix b = (0,0,0,.33,.33,.33)
runmlwin attain cons vrq female, ///
level3(sid: cons) ///
level2(pid: cons) ///
level1(pupil: cons) ///
mcmc(cc) initsb(b) ///
nopause
runmlwin, zratio
where the associated output is ...
Code: Select all
. use http://www.bristol.ac.uk/cmm/media/runmlwin/xc, clear
.
. rename sex female
.
. matrix b = (0,0,0,.33,.33,.33)
.
. runmlwin attain cons vrq female, ///
> level3(sid: cons) ///
> level2(pid: cons) ///
> level1(pupil: cons) ///
> mcmc(cc) initsb(b) ///
> nopause
MLwiN 2.25 multilevel model Number of obs = 3435
Normal response model
Estimation algorithm: MCMC
-----------------------------------------------------------
| No. of Observations per Group
Level Variable | Groups Minimum Average Maximum
----------------+------------------------------------------
sid | 19 92 180.8 290
pid | 148 1 23.2 72
-----------------------------------------------------------
Burnin = 500
Chain = 5000
Thinning = 1
Run time (seconds) = 9.61
Deviance (dbar) = 14722.51
Deviance (thetabar) = 14640.15
Effective no. of pars (pd) = 82.36
Bayesian DIC = 14804.87
------------------------------------------------------------------------------
attain | Mean Std. Dev. ESS P [95% Cred. Interval]
-------------+----------------------------------------------------------------
cons | -10.03677 .2782066 3395 0.000 -10.58556 -9.488309
vrq | .1595779 .00278 4032 0.000 .1542097 .1650367
female | .1153955 .0707506 4990 0.050 -.0219772 .2540834
------------------------------------------------------------------------------
------------------------------------------------------------------------------
Random-effects Parameters | Mean Std. Dev. ESS [95% Cred. Int]
-----------------------------+------------------------------------------------
Level 3: sid |
var(cons) | .0186722 .0226424 341 .0006828 .0824305
-----------------------------+------------------------------------------------
Level 2: pid |
var(cons) | .2793266 .0634311 756 .1687999 .419538
-----------------------------+------------------------------------------------
Level 1: pupil |
var(cons) | 4.257366 .105822 4027 4.056459 4.472385
------------------------------------------------------------------------------
.
. runmlwin, zratio
MLwiN 2.25 multilevel model Number of obs = 3435
Normal response model
Estimation algorithm: MCMC
-----------------------------------------------------------
| No. of Observations per Group
Level Variable | Groups Minimum Average Maximum
----------------+------------------------------------------
sid | 19 92 180.8 290
pid | 148 1 23.2 72
-----------------------------------------------------------
Burnin = 500
Chain = 5000
Thinning = 1
Run time (seconds) = 9.61
Deviance (dbar) = 14722.51
Deviance (thetabar) = 14640.15
Effective no. of pars (pd) = 82.36
Bayesian DIC = 14804.87
------------------------------------------------------------------------------
attain | Mean Std. Dev. z P>|z| [95% Cred. Interval]
-------------+----------------------------------------------------------------
cons | -10.03677 .2782066 -36.08 0.000 -10.58556 -9.488309
vrq | .1595779 .00278 57.40 0.000 .1542097 .1650367
female | .1153955 .0707506 1.63 0.103 -.0219772 .2540834
------------------------------------------------------------------------------
------------------------------------------------------------------------------
Random-effects Parameters | Mean Std. Dev. ESS [95% Cred. Int]
-----------------------------+------------------------------------------------
Level 3: sid |
var(cons) | .0186722 .0226424 341 .0006828 .0824305
-----------------------------+------------------------------------------------
Level 2: pid |
var(cons) | .2793266 .0634311 756 .1687999 .419538
-----------------------------+------------------------------------------------
Level 1: pupil |
var(cons) | 4.257366 .105822 4027 4.056459 4.472385
------------------------------------------------------------------------------
Best wishes
George
Re: Getting p-values and z-ratios after MCMC
Posted: Tue May 22, 2012 12:43 pm
by AnjaScheiwe
Dear George,
many thanks for the quick reply and the explanation re standard error/standard deviation.
I do have version 2.25 installed but the "runmlwin, zratio" command still doesn't return p values for me. I pasted my output and code below:
Code: Select all
. runmlwin maths7sas cons , level3 (dlsoa:cons) level2(s4schoolid: cons) level1(mcsid: cons) mcmc (cc )
> initsb (b) nopause
MLwiN 2.25 multilevel model Number of obs = 8974
Normal response model
Estimation algorithm: MCMC
No. of Observations per Group
Level Variable Groups Minimum Average Maximum
dlsoa 4296 1 2.1 30
s4schoolid 3705 1 2.4 23
maths7sas Mean Std. Dev. z ESS [95% Cred. Interval]
cons 99.35482 .1970295 504.26 1899 98.97349 99.75081
Random-effects Parameters Mean Std. Dev. ESS [95% Cred. Int]
Level 3: dlsoa
var(cons) 19.45014 3.053055 45 13.86732 25.9982
Level 2: s4schoolid
var(cons) 27.89111 3.278613 70 21.74235 34.28221
Level 1: mcsid
var(cons) 185.9832 3.752668 523 178.8243 193.2765
. runmlwin, zratio
MLwiN 2.25 multilevel model Number of obs = 8974
Normal response model
Estimation algorithm: MCMC
No. of Observations per Group
Level Variable Groups Minimum Average Maximum
dlsoa 4296 1 2.1 30
s4schoolid 3705 1 2.4 23
maths7sas Mean Std. Dev. z ESS [95% Cred. Interval]
cons 99.35482 .1970295 504.26 1899 98.97349 99.75081
Random-effects Parameters Mean Std. Dev. ESS [95% Cred. Int]
Level 3: dlsoa
var(cons) 19.45014 3.053055 45 13.86732 25.9982
Level 2: s4schoolid
var(cons) 27.89111 3.278613 70 21.74235 34.28221
Level 1: mcsid
var(cons) 185.9832 3.752668 523 178.8243 193.2765
Did I miss something?
Many thanks!
Anja
Re: Getting p-values and z-ratios after MCMC
Posted: Tue May 22, 2012 12:57 pm
by GeorgeLeckie
Hi Anja,
I see that you are using MLwiN 2.25, but are you sure that you are using the latest version of runmlwin? The runmlwin output still looks like an old version of runmlwin.
Please reinstall runmlwin by typing the following from within Stata...
Once you have done this, please check that the latest version of runmlwin has installed properly by typing...
You should get output similar to the following...
Code: Select all
. which runmlwin
Q:\C-Modelling\runmlwin\development version\runmlwin\runmlwin.ado
*! runmlwin.ado, George Leckie and Chris Charlton, 01Apr2012
Importantly, you should see the date 01Apr2012 in this output.
Best wishes
George
Re: Getting p-values and z-ratios after MCMC
Posted: Tue May 22, 2012 1:09 pm
by AnjaScheiwe
Dear George,
you are right of course - all is working now.
Many thanks and best wishes
Anja