# Re: st: AW: why is this graph all over the place?

 From Nirina F <[email protected]> To [email protected] Subject Re: st: AW: why is this graph all over the place? Date Mon, 21 Sep 2009 16:44:15 -0400

```No you are right, it is in  [1,2].
I put it back to [1,2].
But I don't understand why I don't get the value of the marginal
effects. all I get is:
tab marg

marg |      Freq.     Percent        Cum.
------------+-----------------------------------
0 |      4,119      100.00      100.00
------------+-----------------------------------
Total |      4,119      100.00

On Mon, Sep 21, 2009 at 4:33 PM, Martin Weiss <[email protected]> wrote:
>
> <>
>
>
> List the matrix and come back with the result... Is "varmain" really in
> column 57?
>
>
>
> HTH
> Martin
>
>
> -----Ursprüngliche Nachricht-----
> Von: [email protected]
> [mailto:[email protected]] Im Auftrag von Nirina F
> Gesendet: Montag, 21. September 2009 22:33
> An: [email protected]
> Betreff: Re: st: AW: why is this graph all over the place?
>
> Oh, because When I did
> mat list e(Xmfx_dydx)
> I got:
> e(Xmfx_dydx) [1, 57]
>
>
>
> On Mon, Sep 21, 2009 at 4:26 PM, Martin Weiss <[email protected]> wrote:
>>
>> <>
>>
>> " I corrected the matrix to
>> replace marg=A[1,57] in `i'"
>>
>> Just give me a hint why you did that...
>>
>> You could simply pick one of the values of your varmain, and plug it into
>> -mfx, such as seen here for the value you mentioned earlier. Then list the
>> contents of the matrix with the marginal effects.
>>
>>
>> *************
>> mfx, var(varmain) at(varmain=42.6)
>> mat list e(Xmfx_dydx)
>> *************
>>
>> and show us the output...
>>
>>
>> HTH
>> Martin
>>
>>
>> -----Ursprüngliche Nachricht-----
>> Von: [email protected]
>> [mailto:[email protected]] Im Auftrag von Nirina F
>> Gesendet: Montag, 21. September 2009 22:14
>> An: [email protected]
>> Betreff: Re: st: AW: why is this graph all over the place?
>>
>> Replacing it with -levelsof- made it work! Thanks.
>> But I don't know why the marginal effects are not recorded. However, I
>> corrected the matrix to
>> replace marg=A[1,57] in `i'
>> Nirina
>> On Mon, Sep 21, 2009 at 4:05 PM, Martin Weiss <[email protected]>
> wrote:
>>>
>>> <>
>>>
>>> See, just when I thought I knew Stata: -levels- is not the abbreviation
>> for
>>> -levelsof-, it is an out-of-date command, apparently with limited
>>> functionality. I never noticed, but you did with your non-integer
> value...
>>> So replace it by -levelsof- and it should work:
>>>
>>>
>>> *************
>>> clear*
>>> inp x
>>> 1
>>> 3
>>> 4.5
>>> 4
>>> end
>>> levelsof x
>>> levels x
>>> *************
>>>
>>>
>>>
>>> HTH
>>> Martin
>>>
>>> -----Ursprüngliche Nachricht-----
>>> Von: [email protected]
>>> [mailto:[email protected]] Im Auftrag von Nirina F
>>> Gesendet: Montag, 21. September 2009 21:33
>>> An: [email protected]
>>> Betreff: Re: st: AW: why is this graph all over the place?
>>>
>>> Thank you very much Martin.
>>> I tried the example and it worked perfectly. When I tried with my
>>> variables, the one that replaces mpg has one value that  is not an
>>> integer but is 42.6, therefore I got an error message:
>>> levels varmain
>>> varmain contains non-integer values
>>> r(459);
>>> I guess I should change the command -levels- with another one that
>>> could list the values of varmain.
>>> Also, I tried to recode varmain (42.6=43) but then I don't get
>>> anything inside the new variable marg that keeps the marginal effects.
>>> I am sure something is not right there  because your example works
>>> perfectly. I am going through each  line to check.
>>>
>>> Nirina
>>>
>>> On Thu, Sep 17, 2009 at 6:19 PM, Martin Weiss <[email protected]>
>> wrote:
>>>>
>>>> <>
>>>>
>>>> Try this and check it very carefully. I could not find such a picture in
>>>> Long and Freese (2006),
> http://www.stata-press.com/books/regmodcdvs.html,
>>> -
>>>> which I regard as a bad sign.
>>>>
>>>> Also note I am using the old -mfx- here, as I have not fully gotten my
>>>> round the -margins- command. Hence the -version- statement...
>>>>
>>>>
>>>> *************
>>>> vers 10.1
>>>>
>>>> qui{
>>>>        sysuse auto, clear
>>>>        prob for price mpg rep78 /*
>>>>
>>>>        ins mpg
>>>>        loc uniq= r(N_unique)
>>>>
>>>>        su mpg, mean
>>>>        prgen mpg, from(`r(min)')  /*
>>>>        */ to(`r(max)') /*
>>>>        */ gen(prob) n(`uniq')
>>>>
>>>>
>>>>        levels mpg
>>>>
>>>>        loc i 1
>>>>        gen marg=0
>>>>
>>>>        foreach level in `r(levels)'{
>>>>                mfx, var(mpg) at(mpg=`level')
>>>>                mat A = e(Xmfx_dydx)
>>>>                replace marg=A[1,2] in `i'
>>>>                loc ++i
>>>>        }
>>>>
>>>> }
>>>>
>>>>        tw (connected probp1 probx), /*
>>>>        */ xti("Mpg") ytit("Probability") /*
>>>>        */ name(prob, replace) nodraw
>>>>
>>>>        tw (connected marg probx), /*
>>>>        */ xti("Mpg") ytit("Marginal Effect") /*
>>>>        */ name(marginal, replace) nodraw
>>>>
>>>>
>>>> gr combine prob marginal
>>>> *************
>>>>
>>>>
>>>>
>>>> HTH
>>>> Martin
>>>>
>>>>
>>>> -----Ursprüngliche Nachricht-----
>>>> Von: [email protected]
>>>> [mailto:[email protected]] Im Auftrag von Nirina F
>>>> Gesendet: Donnerstag, 17. September 2009 18:46
>>>> An: [email protected]
>>>> Betreff: Re: st: AW: why is this graph all over the place?
>>>>
>>>> Dear Martin,
>>>>
>>>> Thank you very much for the clarification.
>>>> How about graphing the marginal effects next to that graph?
>>>> nrina
>>>>
>>>> On Thu, Sep 17, 2009 at 12:04 PM, Martin Weiss <[email protected]>
>>> wrote:
>>>>>
>>>>> <>
>>>>>
>>>>> The -prgen- command holds all other covariates at their mean by
> default,
>>>> so
>>>>> my code has already taken care of your concern. To change the -title-s:
>>>>>
>>>>> *************
>>>>> sysuse auto, clear
>>>>> prob for price mpg rep78 /*
>>>>>
>>>>> ins mpg
>>>>> loc uniq= r(N_unique)
>>>>>
>>>>> su mpg, mean
>>>>> prgen mpg, from(`r(min)')  /*
>>>>> */ to(`r(max)') /*
>>>>> */ gen(prob) n(`uniq')
>>>>>
>>>>> tw (connected probp1 probx), /*
>>>>> */ xti("Mpg") ytit("Probability")
>>>>> *************
>>>>>
>>>>>
>>>>>
>>>>> HTH
>>>>> Martin
>>>>>
>>>>>
>>>>> -----Ursprüngliche Nachricht-----
>>>>> Von: [email protected]
>>>>> [mailto:[email protected]] Im Auftrag von Nirina F
>>>>> Gesendet: Donnerstag, 17. September 2009 17:55
>>>>> An: [email protected]
>>>>> Betreff: Re: st: AW: why is this graph all over the place?
>>>>>
>>>>> Thank you very much Martin for  your help. It worked so well.
>>>>>
>>>>> I have a few questions:
>>>>>
>>>>> is it possible also to graph the marginal effects of just the mpg at
>>>>> each value of mpg. I know if you calculate
>>>>> . mfx
>>>>> after
>>>>> . prob for price mpg rep78 headroom trunk weight
>>>>>
>>>>> then you get the marg. effects at the mean value but I would like the
>>>>> marginal effects at each value of mpg but keeping the other variables
>>>>> at their mean. I am aware of the -at- option but it's just difficult
>>>>> to graph it.
>>>>>
>>>>> Also, How would I change the name of the xaxis and yaxis of the
> previous
>>>>> graph?
>>>>>
>>>>> Thank you very much,
>>>>>
>>>>> Nrina
>>>>>
>>>>>
>>>>> On Thu, Sep 17, 2009 at 8:00 AM, Martin Weiss <[email protected]>
>>>> wrote:
>>>>>>
>>>>>> <>
>>>>>>
>>>>>> For these purposes, Scott Long`s -findit spost- comes in handy. Of
>>>> course,
>>>>>> Long and Freese (2006),
>>> http://www.stata-press.com/books/regmodcdvs.html,
>>>>> is
>>>>>> the canonical reference for what you are trying to achieve. Try this:
>>>>>>
>>>>>>
>>>>>> *************
>>>>>> sysuse auto, clear
>>>>>> prob for price mpg rep78 /*
>>>>>>
>>>>>> ins mpg
>>>>>> loc uniq= r(N_unique)
>>>>>>
>>>>>> su mpg, mean
>>>>>> prgen mpg, from(`r(min)')  /*
>>>>>> */ to(`r(max)') /*
>>>>>> */ gen(prob) n(`uniq')
>>>>>>
>>>>>> tw (connected probp1 probx)
>>>>>> *************
>>>>>>
>>>>>>
>>>>>>
>>>>>> HTH
>>>>>> Martin
>>>>>>
>>>>>>
>>>>>> -----Ursprüngliche Nachricht-----
>>>>>> Von: [email protected]
>>>>>> [mailto:[email protected]] Im Auftrag von Nirina F
>>>>>> Gesendet: Donnerstag, 17. September 2009 13:41
>>>>>> An: [email protected]
>>>>>> Betreff: st: why is this graph all over the place?
>>>>>>
>>>>>> Dear all,
>>>>>>
>>>>>> I would like to graph the predicted probabilities of  having foreign
>>>>>> against mpg in order to highlight the slope(basically the marginal
>>>>>> effects) at each value of mpg.
>>>>>> I run the following but it is all over the place because I think it is
>>>>>> giving me the predicted probabilities from different variables.
>>>>>> Could you help me in this please. also if you could help me graph the
>>>>>> marginal effects of mpg from this probit equation against each value
>>>>>> of mpg.
>>>>>>
>>>>>> probit  foreign price mpg rep78 headroom trunk weight
>>>>>> predict pmpg, p
>>>>>> sort mpg
>>>>>> twoway line  pmpg mpg
>>>>>>
>>>>>> Thank you veyr much,
>>>>>> N
>>>>>> *
>>>>>> *   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/
>>>>>
>>>>>
>>>>> *
>>>>> *   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/
>>>>
>>>
>>> *
>>> *   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/
>>
>>
>> *
>> *   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/
>

*
*   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/
```