Problem of using xtologit

Sabyasachi Tripathi

Join Date: Nov 2019

Posts: 20
#1

Problem of using xtologit

26 Jun 2022, 07:53

Dear Professors,
I am using xtologit for the order logit model. I have four income group countries. Income variable specified as
1 = low income group
2 =Lower middle income
3 = Upper middle income
4 = high income
I am trying to asses the impact of urban population on fertility rate

So I run the following
xtset income
xtologit fertilityrate urbanpopulation

Then I get the following results

Fitting comparison model:

J(): 3900 unable to allocate real <tmp>[6660,8935878]
_gsem_eval_ordinal(): - function returned error
_gsem_eval_iid__obs(): - function returned error
_gsem_eval_iid__wrk(): - function returned error
_gsem_eval_iid(): - function returned error
mopt__calluser_v(): - function returned error
opt__eval_nr_v2(): - function returned error
opt__eval(): - function returned error
opt__looputil_iter0_common(): - function returned error
opt__looputil_iter0_nr(): - function returned error
opt__loop_nr(): - function returned error
opt__loop(): - function returned error
_moptimize(): - function returned error
_gsem_start__fixed(): - function returned error
_gsem_start_mecmd(): - function returned error
_gsem_build(): - function returned error
_gsem_parse(): - function returned error
st_gsem_parse(): - function returned error
<istmt>: - function returned error
r(3900);

Could you please help me how to solve this problem.

Thanking you,

With Kind Regards,
Sabyasachi Tripathi
Tags: None
Richard Williams

Join Date: Apr 2014

Posts: 5043
#2

26 Jun 2022, 08:41

Xtsetting on a variable with only 4 possible values doesn’t make much sense. Do you really have panel data, and if so what is the panelid?

Also how is your dependent variable coded? Is it really ordinal?

-------------------------------------------
Richard Williams
Professor Emeritus of Sociology
University of Notre Dame
StataNow Version: 19.5 MP (2 processor)
EMAIL: [email protected]
WWW: https://academicweb.nd.edu/~rwilliam/
Comment
Sabyasachi Tripathi

Join Date: Nov 2019

Posts: 20
#3

26 Jun 2022, 08:56

Dear Prof. Richard Williams
Thank you for your reply. Yes I have panel data from 1960 to 2021 and country level data on fertility rate and total urban population. My question is that is there any variation in the results based on their income level. So I coded countries by 4 different orders as explained before.
I followed the manuals from the following STATA manuals...they also have given similar example.
https://www.stata.com/manuals13/xtxtologit.pdf

Thank you for your kind help.
Comment
William Lisowski

Join Date: Dec 2014

Posts: 10150
#4

26 Jun 2022, 09:16

Please show us the output of

Code:

codebook fertilityrate urbanpopulation income

To assure maximum readability of results that you post, please copy them from the Results window into a code block in the Forum editor using code delimiters [CODE] and [/CODE], as explained in section 12 of the Statalist FAQ linked to at the top of the page. For example, the following difficlut-to-read text

[CODE]
. codebook price weight foreign

------------------------------------------------------------------------------------------------
price Price
------------------------------------------------------------------------------------------------

Type: Numeric (int)

Range: [3291,15906] Units: 1
Unique values: 74 Missing .: 0/74

Mean: 6165.26
Std. dev.: 2949.5

Percentiles: 10% 25% 50% 75% 90%
3895 4195 5006.5 6342 11385

------------------------------------------------------------------------------------------------
weight Weight (lbs.)
------------------------------------------------------------------------------------------------

Type: Numeric (int)

Range: [1760,4840] Units: 10
Unique values: 64 Missing .: 0/74

Mean: 3019.46
Std. dev.: 777.194

Percentiles: 10% 25% 50% 75% 90%
2020 2240 3190 3600 4060

------------------------------------------------------------------------------------------------
foreign Car origin
------------------------------------------------------------------------------------------------

Type: Numeric (byte)
Label: origin

Range: [0,1] Units: 1
Unique values: 2 Missing .: 0/74

Tabulation: Freq. Numeric Label
52 0 Domestic
22 1 Foreign
[/CODE]

will be presented in the post as the following, using a more appropriate font

Code:

. codebook price weight foreign ------------------------------------------------------------------------------------------------ price Price ------------------------------------------------------------------------------------------------ Type: Numeric (int) Range: [3291,15906] Units: 1 Unique values: 74 Missing .: 0/74 Mean: 6165.26 Std. dev.: 2949.5 Percentiles: 10% 25% 50% 75% 90% 3895 4195 5006.5 6342 11385 ------------------------------------------------------------------------------------------------ weight Weight (lbs.) ------------------------------------------------------------------------------------------------ Type: Numeric (int) Range: [1760,4840] Units: 10 Unique values: 64 Missing .: 0/74 Mean: 3019.46 Std. dev.: 777.194 Percentiles: 10% 25% 50% 75% 90% 2020 2240 3190 3600 4060 ------------------------------------------------------------------------------------------------ foreign Car origin ------------------------------------------------------------------------------------------------ Type: Numeric (byte) Label: origin Range: [0,1] Units: 1 Unique values: 2 Missing .: 0/74 Tabulation: Freq. Numeric Label 52 0 Domestic 22 1 Foreign

Compare this to the code you posted in your previous Statalist topic at

https://www.statalist.org/forums/for...69#post1636469

which received no reponse, perhaps because it was difficult to read.
1 like
Comment
Richard Williams

Join Date: Apr 2014

Posts: 5043
#5

26 Jun 2022, 09:40

You may want to include income in the model but you do not want to xtset on it. You probably want something like

xtset country year

-------------------------------------------
Richard Williams
Professor Emeritus of Sociology
University of Notre Dame
StataNow Version: 19.5 MP (2 processor)
EMAIL: [email protected]
WWW: https://academicweb.nd.edu/~rwilliam/
1 like
Comment
Sabyasachi Tripathi

Join Date: Nov 2019

Posts: 20
#6

26 Jun 2022, 11:16

Dear Prof. Richard Williams

Yes now I used xtset country year
Now my dependent variable is income group variable "income_group" and independent variable is urban population (urban_p). Then I have ended up with the following problem:

xtologit income_group ln_urban_p

Fitting comparison model:

Iteration 0: log likelihood = -15939.176
Iteration 1: log likelihood = -15934.806
Iteration 2: log likelihood = -15934.806

Refining starting values:

Grid node 0: log likelihood = -8759.1347

Fitting full model:

Iteration 0: log likelihood = -8759.1347
Iteration 1: log likelihood = -1501.5138
Iteration 2: log likelihood = -1205.6597
Iteration 3: log likelihood = -625.09707 (not concave)
Iteration 4: log likelihood = -552.9766 (not concave)
Iteration 5: log likelihood = -495.75167 (not concave)
cannot compute an improvement -- discontinuous region encountered
r(430);

Could you please help me.

With Kind Regards,
Comment
Richard Williams

Join Date: Apr 2014

Posts: 5043
#7

26 Jun 2022, 11:45

Info like William requested may help.

-------------------------------------------
Richard Williams
Professor Emeritus of Sociology
University of Notre Dame
StataNow Version: 19.5 MP (2 processor)
EMAIL: [email protected]
WWW: https://academicweb.nd.edu/~rwilliam/
Comment

Andrew Musau

Join Date: Oct 2014
Posts: 10482

26 Jun 2022, 13:57

As others have commented, you need to provide more information relating to your data. You have asked about gllamm from SSC (via email). Here is the equivalent syntax:

Code:

webuse tvsfpors, clear
xtset school
*XTOLOGIT
xtologit thk prethk cc##tv, nolog

*ssc install gllamm
*GLLAMM
gen cctv= c.cc#c.tv
gllamm thk prethk cc tv cctv,  link(ologit) fam(binom) i(school) nolog

Res.:

Code:

. xtologit thk prethk cc##tv, nolog

Random-effects ordered logistic regression      Number of obs     =      1,600
Group variable: school                          Number of groups  =         28

Random effects u_i ~ Gaussian                   Obs per group:
                                                              min =         18
                                                              avg =       57.1
                                                              max =        137

Integration method: mvaghermite                 Integration pts.  =         12

                                                Wald chi2(4)      =     128.06
Log likelihood  = -2119.7428                    Prob > chi2       =     0.0000

------------------------------------------------------------------------------
         thk |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
      prethk |   .4032892     .03886    10.38   0.000      .327125    .4794534
        1.cc |   .9237904    .204074     4.53   0.000     .5238127    1.323768
        1.tv |   .2749937   .1977424     1.39   0.164    -.1125744    .6625618
             |
       cc#tv |
        1 1  |  -.4659256   .2845963    -1.64   0.102    -1.023724    .0918728
-------------+----------------------------------------------------------------
       /cut1 |  -.0884493   .1641062                     -.4100916     .233193
       /cut2 |   1.153364    .165616                      .8287625    1.477965
       /cut3 |    2.33195   .1734199                      1.992053    2.671846
-------------+----------------------------------------------------------------
   /sigma2_u |   .0735112   .0383106                      .0264695    .2041551
------------------------------------------------------------------------------
LR test vs. ologit model: chibar2(01) = 10.72         Prob >= chibar2 = 0.0005

.
.
.
. *ssc install gllamm

.
. *GLLAMM

.
. gen cctv= c.cc#c.tv

.
. gllamm thk prethk cc tv cctv,  link(ologit) fam(binom) i(school) nolog
 
number of level 1 units = 1600
number of level 2 units = 28
 
Condition Number = 16.909052
 
gllamm model
 
log likelihood = -2119.7554
 
------------------------------------------------------------------------------
         thk |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
thk          |
      prethk |   .4034971   .0388353    10.39   0.000     .3273814    .4796128
          cc |   .9099342   .2070142     4.40   0.000     .5041938    1.315675
          tv |   .2679143   .2018645     1.33   0.184    -.1277329    .6635615
        cctv |  -.4567022   .2853622    -1.60   0.110    -1.016002    .1025974
-------------+----------------------------------------------------------------
_cut11       |
       _cons |  -.0964273   .1702289    -0.57   0.571    -.4300698    .2372152
-------------+----------------------------------------------------------------
_cut12       |
       _cons |   1.145341   .1713372     6.68   0.000     .8095264    1.481156
-------------+----------------------------------------------------------------
_cut13       |
       _cons |   2.323844   .1785075    13.02   0.000     1.973976    2.673712
------------------------------------------------------------------------------
 
 
Variances and covariances of random effects
------------------------------------------------------------------------------

 
***level 2 (school)
 
    var(1): .07305962 (.03834715)
------------------------------------------------------------------------------

To request for adaptive quadrature instead of Newton-Raphson be used in the maximization, add the option -adapt-. In your case, the syntax is:

Code:

gllamm income_group ln_urban_p, link(ologit) fam(binom) i(country) adapt

This does not guarantee that you will get a result, you may have data problems. So follow the advice you have been given in this thread if you still face problems.

Comment

Sabyasachi Tripathi

Join Date: Nov 2019

Posts: 20
#9

26 Jun 2022, 22:37

Please see the data...

codebook ln_fertility ln_urban_p income_group

---------------------------------------------------------------------------------------------------------------------------------------------
ln_fertility (unlabeled)
---------------------------------------------------------------------------------------------------------------------------------------------

type: numeric (float)

range: [-.17793119,2.1355855] units: 1.000e-09
unique values: 3,538 missing .: 147/6,660

mean: 1.23117
std. dev: .534645

percentiles: 10% 25% 50% 75% 90%
.500775 .738598 1.27508 1.73853 1.90405

---------------------------------------------------------------------------------------------------------------------------------------------
ln_urban_p (unlabeled)
---------------------------------------------------------------------------------------------------------------------------------------------

type: numeric (float)

range: [6.7753663,19.993414] units: 1.000e-07
unique values: 6,655 missing .: 0/6,660

mean: 14.628
std. dev: 2.23151

percentiles: 10% 25% 50% 75% 90%
11.2924 13.3264 14.9921 16.0573 17.3091

---------------------------------------------------------------------------------------------------------------------------------------------
income_group (unlabeled)
---------------------------------------------------------------------------------------------------------------------------------------------

type: numeric (float)

range: [1,4] units: 1
unique values: 4 missing .: 0/6,660

tabulation: Freq. Value
720 1
1,860 2
2,940 3
1,140 4
Comment
Sabyasachi Tripathi

Join Date: Nov 2019

Posts: 20
#10

26 Jun 2022, 23:22

codebook urban_p income_group fertility

urban_p URBAN_P

type: numeric (long)

range: [876,4.820e+08] units: 1
unique values: 6,655 missing .: 0/6,660

mean: 1.4e+07
std. dev: 3.6e+07

percentiles: 10% 25% 50% 75% 90%
80209.5 613178 3.2e+06 9.4e+06 3.3e+07

income_group (unlabeled)

type: numeric (float)

range: [1,4] units: 1
unique values: 4 missing .: 0/6,660

tabulation: Freq. Value
720 1
1,860 2
2,940 3
1,140 4

fertility FERTILITY

type: numeric (float)

range: [.837,8.462] units: .0001
unique values: 3,538 missing .: 147/6,660

mean: 3.91933
std. dev: 1.9357

percentiles: 10% 25% 50% 75% 90%
1.65 2.093 3.579 5.689 6.713
Comment
Sabyasachi Tripathi

Join Date: Nov 2019

Posts: 20
#11

26 Jun 2022, 23:32

Now, I found the following problem
gllamm income_group ln_urban_p, link(ologit) fam(binom) i(country1) adapt

Running adaptive quadrature
Iteration 0: log likelihood = -1.3176213
(error occurred in ML computation)
(use trace option and check correctness of initial model)

.
Comment
Andrew Musau

Join Date: Oct 2014

Posts: 10482
#12

27 Jun 2022, 04:58

type: numeric (long)

range: [876,4.820e+08] units: 1
unique values: 6,655 missing .: 0/6,660

mean: 1.4e+07
std. dev: 3.6e+07

percentiles: 10% 25% 50% 75% 90%
80209.5 613178 3.2e+06 9.4e+06 3.3e+07

I think it has to do with the very large values of this variable. I will simulate some data and get back to you.
2 likes
Comment
Richard Williams

Join Date: Apr 2014

Posts: 5043
#13

27 Jun 2022, 10:58

Some tips for dealing with convergence problems are at

https://www3.nd.edu/~rwilliam/xsoc73994/L02.pdf

I agree with Andrew — rescaling the variable with the huge values (e.g. divide by 1 million) May help.

-------------------------------------------
Richard Williams
Professor Emeritus of Sociology
University of Notre Dame
StataNow Version: 19.5 MP (2 processor)
EMAIL: [email protected]
WWW: https://academicweb.nd.edu/~rwilliam/
2 likes
Comment

Announcement