Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: st: Prediction for fractional logit
From
Nick Cox <[email protected]>
To
[email protected]
Subject
Re: st: Prediction for fractional logit
Date
Tue, 31 May 2011 13:42:16 +0100
If you insist on using the -xb- option with -predict-, that is what
you get! So, just go for the default.
Nick
On Tue, May 31, 2011 at 1:33 PM, S N <[email protected]> wrote:
> Nick and Maarten,
>
> I am pasting the code and the results below. Thanks. Shonda
>
>
> Variable | Obs Mean Std. Dev. Min Max
> -------------+--------------------------------------------------------
> prop_gra | 53 .5790939 .4135732 0 1
> prop_obc | 54 .160307 .2915177 0 1
> prop_p1 | 54 .2106382 .2555498 0 .9690722
> land1 | 54 .4508955 .3359333 0 1
> land2 | 54 .1073159 .1590588 0 .7647059
> prop_cash | 53 16.5283 24.05011 0 80
> estdc_land1 | 54 .1926632 .3186031 0 .9230769
>
> estd_ext | Freq. Percent Cum.
> ------------+-----------------------------------
> 0 | 36 65.45 65.45
> 1 | 19 34.55 100.00
> ------------+-----------------------------------
>
> estd_comm | Freq. Percent Cum.
> ------------+-----------------------------------
> 0 | 32 58.18 58.18
> 1 | 23 41.82 100.00
> ------------+-----------------------------------
>
>
> prop_obc, prop_p1, prop_land1, prop_land2 are all proportions
> themselves. estd_ext, estd_comm, d are dummies, estdc_land1 is an
> interaction term between prop_land1 and estd_comm (so interaction with
> a dummy and a proportion).
>
> ****
>
> glm prop_gra prop_obc prop_p1 prop_land1 prop_land2 prop_cash estd_ext
> estd_comm estdc_land1 d, family (binomial) link (logit) nolog
>
>
> Generalized linear models No. of obs = 52
> Optimization : ML Residual df
> = 42
>
> Scale parameter = 1
> Deviance = 24.98012225 (1/df) Deviance = .5947648
> Pearson = 25.64304946 (1/df) Pearson = .6105488
>
> Variance function: V(u) = u*(1-u/1) [Binomial]
> Link function : g(u) = ln(u/(1-u)) [Logit]
>
>
> AIC = 1.20732
> Log likelihood = -21.39033265 BIC = -140.9721
>
> ------------------------------------------------------------------------------
> | OIM
> prop_gra | Coef. Std. Err. z P>|z| [95%
> Conf. Interval]
> -------------+----------------------------------------------------------------
> prop_obc | 1.20795 1.521787 0.79 0.427 -1.774697 4.190597
> prop_p1 | 2.271956 1.642525 1.38 0.167 -.947333 5.491246
> prop_land1 | 1.152989 1.597026 0.72 0.470 -1.977125 4.283103
> prop_land2 | 1.688074 2.455906 0.69 0.492 -3.125414 6.501561
> prop_cash | .0509537 .0253546 2.01 0.044 .0012596 .1006478
> estd_ext | 1.498499 1.0195 1.47 0.142 -.4996845 3.496682
> estd_comm | 2.725557 1.472281 1.85 0.064 -.1600611 5.611175
> estdc_land1 | -3.777789 2.618245 -1.44 0.149 -8.909454 1.353876
> d | -.7853135 .9559989 -0.82 0.411
> -2.659037 1.08841
> _cons | -2.317875 1.36365 -1.70 0.089 -4.990581
> .3548302
>
> ****
> predict gra_pred, xb
>
> *****
> summarize gra_pred
>
> Variable | Obs Mean Std. Dev. Min Max
> -------------+--------------------------------------------------------
> gra_pred | 52 .5611184 1.595445 -2.67838 3.891145
>
> *******
>
>
> On Tue, May 31, 2011 at 8:09 AM, Nick Cox <[email protected]> wrote:
>> I agree with your implication that this should not happen. Please tell
>> us more about what you did, including exact -glm- and -predict-
>> commands and -summarize- results for all variables in the model.
>>
>> Nick
>>
>> On Tue, May 31, 2011 at 1:04 PM, S N <[email protected]> wrote:
>>
>>> I am using glm in combination with the link(logit) family(binomial)
>>> robust options to estimate proportion [0,1] of households engaged in a
>>> specific activity. However, the predict command thereafter provides me
>>> with predicted values that takes the values outside [0,1]. What could
>>> I be doing wrong?
*
* 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/