st: Maximum likelihood simulation

classic Classic list List threaded Threaded
6 messages Options
Reply | Threaded
Open this post in threaded view
|

st: Maximum likelihood simulation

Katia Bobulova
Dear All,

I am still trying to perform my maximum likelihood. The point is that
I receive this error message:"could not calculate numerical
derivatives discontinuous region with missing values encountered". I
don't have any value that is equal to zero in my data. So maybe this
is a problem of inconsistency,there could be some values that go to
zero or infinity.

At this point I would like to simulate my maximum likelihood in order
to have initial feasible values.

Following my program for the maximum likelihood:

program define myprogr
version 10.0
args lnf a d b mo mq eo eq n g

tempvar ma md mb mmo mmq meo meq mn mg

quietly gen double `ma'=1-`a'
quietly gen double `md'=1-`d'
quietly gen double `mb'=1-`b'
quietly gen double `mmo'=1-`mo'
quietly gen double `mmq'=1-`mq'
quietly gen double `meo'=1-`eo'
quietly gen double `meq'=1-`eq'
quietly gen double `mn'=1-`n'
quietly gen double `mg'=1-`g'

quietly replace
`lnf'=ln((`a')*(`d')*((((`b')*(`mq')*(`g')+(`b')*(`mmq')*(`eq')*(`n'))^bq)*(((`mb')*(`mo')*(`g')+(`mb')*(`mmo')*(`eo')*(`n'))^bo)*(((`b')*(`mmq')*(`eq')*(`mn'))^sq)*(((`mb')*(`mmo')*(`eo')*(`mn'))^so)*(((`b')*(`mmq')*(`meq')+(`mb')*(`mmo')*(`meo'))^nt))+(`a')*(`md')*((((`b')*(`mmq')*(`eq')*(`n'))^bq)*(((`mb')*(`mmo')*(`eo')*(`n'))^bo)*(((`b')*(`mq')*(`mg')+(`b')*(`mmq')*(`eq')*(`mn'))^sq)*(((`mb')*(`mo')*(`mg')+(`mb')*(`mmo')*(`eo')*(`mn'))^so)*(((`b')*(`mmq')*(`meq')+(`mb')*(`mmo')*(`meo'))^nt))+(`ma')*((((`b')*(`eq')*(`n'))^bq)*(((`mb')*(`eo')*(`n'))^bo)*(((`b')*(`eq')*(`mn'))^sq)*(((`mb')*(`eo')*(`mn'))^so)*(((`b')*(`meq')+(`mb')*(`meo'))^nt))+2.71^(-745))

end


ml model lf myprogr (a:) (d:) (b:)(mo:)(mq:)(eo:)(eq:)(n:)(g:)
ml search
ml maximize, difficult


I would like to have simulated values for a,d,b,mo,mq,eo,eq,n,g

After end would should I write?
simulate....

Could you please help me? I am encountering many difficulties that i
don't know how to overcome.

Thanks a lot

Katia
*
*   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/
Reply | Threaded
Open this post in threaded view
|

st: graphing problem

Mak, Timothy
Hi Statalist,

I'm joining the list again after a period of absence. For some reason, I
don't seem to be able to send directly to
[hidden email], but if I reply to a post, then it seems
to get there. My original message was as follows:

Hi Statalist,

 

Is there any reason why Stata won't let you draw a graph whose y-axis'
range is less than the actual range of the data? Is there a way to get
round this?

 

The motivation comes from doing a graph where there is one group which
has a articularly large y value. If I graph it with the other groups,
then all the other groups will be very small, and their relative heights
not very distinguishable. If I can focus on the other groups and let the
outlying group 'goes off the scale', that's clearly quite a good device
to have. In fact, the problem can't be overcome even if I artificially
set the outlying y value to the maximum of the y range, because the top
will look different, particularly when I want to graph error intervals.

 

Thanks for your help.

 

Tim

*
*   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/
Reply | Threaded
Open this post in threaded view
|

st: RE: graphing problem

Martin Weiss-5

Line for the server...

Try http://www.stata-journal.com/sjpdf.html?articlenum=gr0019


HTH
Martin

-----Original Message-----
From: [hidden email]
[mailto:[hidden email]] On Behalf Of Mak, Timothy
Sent: Friday, December 05, 2008 4:15 PM
To: [hidden email]
Subject: st: graphing problem

Hi Statalist,

I'm joining the list again after a period of absence. For some reason, I
don't seem to be able to send directly to
[hidden email], but if I reply to a post, then it seems
to get there. My original message was as follows:

Hi Statalist,

 

Is there any reason why Stata won't let you draw a graph whose y-axis'
range is less than the actual range of the data? Is there a way to get
round this?

 

The motivation comes from doing a graph where there is one group which
has a articularly large y value. If I graph it with the other groups,
then all the other groups will be very small, and their relative heights
not very distinguishable. If I can focus on the other groups and let the
outlying group 'goes off the scale', that's clearly quite a good device
to have. In fact, the problem can't be overcome even if I artificially
set the outlying y value to the maximum of the y range, because the top
will look different, particularly when I want to graph error intervals.

 

Thanks for your help.

 

Tim

*
*   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/

*
*   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/
Reply | Threaded
Open this post in threaded view
|

st: RE: RE: graphing problem

Mak, Timothy
Hi Martin,

Thanks for the article. However, as the article says, Stata determines
the minimum range of the graph according to the range of the data. It
remains impossible to set the range of the graph to be smaller than that
of the data.

Tim

-----Original Message-----
From: [hidden email]
[mailto:[hidden email]] On Behalf Of Martin Weiss
Sent: 05 December 2008 15:17
To: [hidden email]
Subject: st: RE: graphing problem


Line for the server...

Try http://www.stata-journal.com/sjpdf.html?articlenum=gr0019


HTH
Martin

-----Original Message-----
From: [hidden email]
[mailto:[hidden email]] On Behalf Of Mak, Timothy
Sent: Friday, December 05, 2008 4:15 PM
To: [hidden email]
Subject: st: graphing problem

Hi Statalist,

I'm joining the list again after a period of absence. For some reason, I
don't seem to be able to send directly to
[hidden email], but if I reply to a post, then it seems
to get there. My original message was as follows:

Hi Statalist,

 

Is there any reason why Stata won't let you draw a graph whose y-axis'
range is less than the actual range of the data? Is there a way to get
round this?

 

The motivation comes from doing a graph where there is one group which
has a articularly large y value. If I graph it with the other groups,
then all the other groups will be very small, and their relative heights
not very distinguishable. If I can focus on the other groups and let the
outlying group 'goes off the scale', that's clearly quite a good device
to have. In fact, the problem can't be overcome even if I artificially
set the outlying y value to the maximum of the y range, because the top
will look different, particularly when I want to graph error intervals.

 

Thanks for your help.

 

Tim

*
*   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/

*
*   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/

*
*   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/
Reply | Threaded
Open this post in threaded view
|

Re: st: RE: RE: graphing problem

Scott Merryman
Perhaps the "FAQ: How can I show scale breaks on graphs? "

 http://www.stata.com/support/faqs/graphics/scbreak.html

would be helpful.

Scott


On Fri, Dec 5, 2008 at 10:18 AM, Mak, Timothy
<[hidden email]> wrote:
> Hi Martin,
>
> Thanks for the article. However, as the article says, Stata determines
> the minimum range of the graph according to the range of the data. It
> remains impossible to set the range of the graph to be smaller than that
> of the data.
>
> Tim
*
*   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/
Reply | Threaded
Open this post in threaded view
|

st: RE: RE: RE: graphing problem

Nick Cox
In reply to this post by Mak, Timothy
Stata here as elsewhere applies the principle that a graph should be
honest about your data. Equivalently, it does not want to make it easy
for you to omit data from a graph by accident. Thus if you want to omit
data on purpose you should do on purpose by applying -if-.

An alternative might be to use a log scale.

Nick
[hidden email]

Mak, Timothy

Thanks for the article. However, as the article says, Stata determines
the minimum range of the graph according to the range of the data. It
remains impossible to set the range of the graph to be smaller than that
of the data.

Martin Weiss

Try http://www.stata-journal.com/sjpdf.html?articlenum=gr0019

Mak, Timothy

Is there any reason why Stata won't let you draw a graph whose y-axis'
range is less than the actual range of the data? Is there a way to get
round this?

The motivation comes from doing a graph where there is one group which
has a articularly large y value. If I graph it with the other groups,
then all the other groups will be very small, and their relative heights
not very distinguishable. If I can focus on the other groups and let the
outlying group 'goes off the scale', that's clearly quite a good device
to have. In fact, the problem can't be overcome even if I artificially
set the outlying y value to the maximum of the y range, because the top
will look different, particularly when I want to graph error intervals.

 

*
*   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/