Unordered imputed variable

Welcome to the forum for runmlwin users. Feel free to post your question about runmlwin here. The Centre for Multilevel Modelling take no responsibility for the accuracy of these posts, we are unable to monitor them closely. Do go ahead and post your question and thank you in advance if you find the time to post any answers!

Go to runmlwin: Running MLwiN from within Stata >> http://www.bristol.ac.uk/cmm/software/runmlwin/
Forum rules
Post Reply
likestatistic
Posts: 39
Joined: Fri Sep 12, 2014 4:11 pm

Unordered imputed variable

Post by likestatistic »

Hi All,
I have 2 variables :
var1 continuous with missing values (100)
var2 unordered categorical with missing values (25)
I have imputed the missing values
now I am running the following programme:

xi: mi est, cmdok: runmlwin y cons var1t i.var2 , level2(area: cons,) level1(indiv: cons) nopause

And the result still have missing data corresponding to var2 (25)

How should I do to tell runmlwin to use the imputed values of the unordered categorical variable?

Many thanks

L
Top
GeorgeLeckie
Site Admin
Posts: 432
Joined: Fri Apr 01, 2011 2:14 pm

Re: Unordered imputed variable

Post by GeorgeLeckie »

Dear likestatistic,

Please will you provide an example which works with the regress command. We will then try to work out why the runmlwin command does not work when you substitute it for the regress command.

Best wishes

George
Top
likestatistic
Posts: 39
Joined: Fri Sep 12, 2014 4:11 pm

Re: Unordered imputed variable

Post by likestatistic »

I hope that you will be able to use this table.
It is a sample of my data.
Var1 and var2 had missing values and have been imputed
I want to use them in my multilevel model to predict y
using for example
xi: mi est, cmdok: runmlwin y cons var1 i.var2 , level2(area: cons,) level1(id: cons) nopause

Many Thanks

L
Attachments
data.xls
(16.75 KiB) Downloaded 566 times
Top
GeorgeLeckie
Site Admin
Posts: 432
Joined: Fri Apr 01, 2011 2:14 pm

Re: Unordered imputed variable

Post by GeorgeLeckie »

Hi Likestatistic,

Sorry perhaps I was not clear. In order to try to find out what might not be working with the rumlwin command, we really need you to provide us with a series of commands which work with your data with the regress command. If you are struggling to implement the single-level version of what you want with the regress command then I would suggest you email Stata support directly for help.

Best wishes

George
Top
likestatistic
Posts: 39
Joined: Fri Sep 12, 2014 4:11 pm

Re: Unordered imputed variable

Post by likestatistic »

Hi,
Thank you for the suggestion and sorry to have bothered you. My problem is sorted, a simple mistake in the command and a bug due to the fact that I' am working with more than 300thousand observations .

Here are the results with the data I provided

Code: Select all

*command 
insheet using "N:\data.xls",clear
drop _mi_miss _1_var2 _2_var2 _3_var2 _4_var2 _5_var2 _6_var2 _7_var2 _8_var2 _9_var2 ///
_10_var2 _1_var1 _2_var1 _3_var1 _4_var1 _5_var1 _6_var1 _7_var1 _8_var1 _9_var1 _10_var1
mi set wide
mi describ
mi register imputed var1 var2
mi impute chained (mlogit) var1 (regress) var2= y, add(5) rseed(1523)
*model in stata
mi estimate, : mixed y i.var1 var2 || area:, reml

*model in stata using mlwin
sort area id
gen cons=1

mi est, mcerror cmdok: runmlwin y cons i.var1 var2, level2(area: cons,) level1(id: cons) nopause

*results

. drop _mi_miss _1_var2 _2_var2 _3_var2 _4_var2 _5_var2 _6_var2 _7_var2 _8_var2 
> _9_var2 ///
> _10_var2 _1_var1 _2_var1 _3_var1 _4_var1 _5_var1 _6_var1 _7_var1 _8_var1 _9_va
> r1 _10_var1

. mi set wide

. mi describ

  Style:  wide
          last mi update 01oct2014 18:15:41, 0 seconds ago

  Obs.:   complete          203
          incomplete          0  (M = 0 imputations)
          ---------------------
          total             203

  Vars.:  imputed:  0

          passive:  0

          regular:  0

          system:   1; _mi_miss

         (there are 5 unregistered variables)

. mi register imputed var1 var2

. mi impute chained (mlogit) var1 (regress) var2= y, add(5) rseed(1523)

Conditional models:
              var2: regress var2 i.var1 y
              var1: mlogit var1 var2 y

Performing chained iterations ...

Multivariate imputation                     Imputations =        5
Chained equations                                 added =        5
Imputed: m=1 through m=5                        updated =        0

Initialization: monotone                     Iterations =       50
                                                burn-in =       10

              var1: multinomial logistic regression
              var2: linear regression

------------------------------------------------------------------
                   |               Observations per m             
                   |----------------------------------------------
          Variable |   Complete   Incomplete   Imputed |     Total
-------------------+-----------------------------------+----------
              var1 |        165           38        38 |       203
              var2 |        184           19        19 |       203
------------------------------------------------------------------
(complete + incomplete = total; imputed is the minimum across m
 of the number of filled-in observations.)

. 
. mi estimate, : mixed y i.var1 var2 || area:, reml

Multiple-imputation estimates                   Imputations        =         5
Mixed-effects REML regression                   Number of obs      =       203

Group variable: area                            Number of groups   =        17
                                                Obs per group: min =         1
                                                               avg =      11.9
                                                               max =        25

                                                Average RVI        =    0.1133
                                                Largest FMI        =    0.3464
DF adjustment:   Large sample                   DF:     min        =     40.37
                                                        avg        =  59588.91
                                                        max        = 226426.59
Model F test:       Equal FMI                   F(   3,  186.6)    =      1.39
                                                Prob > F           =    0.2487

------------------------------------------------------------------------------
           y |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
        var1 |
          2  |   162.8015   90.83603     1.79   0.073    -15.39469    340.9978
          4  |   64.80362   313.3858     0.21   0.837     -568.393    698.0002
             |
        var2 |   7.725514   6.676412     1.16   0.248    -5.410592    20.86162
       _cons |   2038.737   1087.449     1.87   0.062    -100.6839    4178.158
------------------------------------------------------------------------------

------------------------------------------------------------------------------
  Random-effects Parameters  |   Estimate   Std. Err.     [95% Conf. Interval]
-----------------------------+------------------------------------------------
area: Identity               |
                   sd(_cons) |   97.24503   63.21491      27.19782    347.6969
-----------------------------+------------------------------------------------
                sd(Residual) |   586.6303   30.30025      530.1495    649.1284
------------------------------------------------------------------------------

. 
. sort area id

. gen cons=1

. 
. do "C:\Users\SOUEDR~1\AppData\Local\Temp\STD00000000.tmp"

. mi est, cmdok: runmlwin y cons i.var1 var2, level2(area: cons,) level1(id: con
> s) nopause

Multiple-imputation estimates                     Imputations     =          5
Normal response model                             Number of obs   =        203
                                                  Average RVI     =     0.1146
                                                  Largest FMI     =     0.3512
DF adjustment:   Large sample                     DF:     min     =      39.30
                                                          avg     =   91844.87
                                                          max     =  424369.54

------------------------------------------------------------------------------
             |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
FP1          |
        cons |   2024.962   1076.085     1.88   0.061     -92.0137    4141.938
     _2_var1 |   162.8681   89.82891     1.81   0.070    -13.35063    339.0868
     _4_var1 |   61.05552   310.9212     0.20   0.845    -567.6872    689.7982
        var2 |   7.804148    6.60805     1.18   0.238    -5.196928    20.80522
-------------+----------------------------------------------------------------
RP2          |
   var(cons) |   7242.816    11863.4     0.61   0.542     -16009.1    30494.73
-------------+----------------------------------------------------------------
RP1          |
   var(cons) |   338825.9   34927.96     9.70   0.000     270367.7    407284.1
------------------------------------------------------------------------------
Note: number of groups varies among imputations.
Note: number of observations per group varies among imputations.
Regards
L
Top
GeorgeLeckie
Site Admin
Posts: 432
Joined: Fri Apr 01, 2011 2:14 pm

Re: Unordered imputed variable

Post by GeorgeLeckie »

Great,
Very please you got it working
Best wishes
George
Top
Post Reply

Return to “runmlwin user forum”