Bookmark and Share

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]

st: Big Differences of Significance in Marginal Effect Estimation using margeff


From   "[email protected]" <[email protected]>
To   [email protected]
Subject   st: Big Differences of Significance in Marginal Effect Estimation using margeff
Date   Wed, 18 May 2011 16:07:21 -0400

Hi

I am running Poisson regression.  The following is the result of Poisson

regression with beta coefficients



. poisson y x1 x2 x3 x4 x5 x6 x7 x8 x9 x10 x11

Iteration 0:   log likelihood = -55.606422

Iteration 1:   log likelihood = -55.422123

Iteration 2:   log likelihood =  -55.42153

Iteration 3:   log likelihood =  -55.42153

Poisson regression                                Number of obs   =

55

                                                 LR chi2(11)     =

66.47

                                                 Prob > chi2     =

0.0000

Log likelihood =  -55.42153                       Pseudo R2       =

0.3749

------------------------------------------------------------------------------

        y   |      Coef.   Std. Err.      z    P>|z|     [95% Conf.

Interval]

-------------+----------------------------------------------------------------

       x1   |   .0893232   .1247445     0.72   0.474    -.1551715

.3338179

       x2   |   -.092306   .1944331    -0.47   0.635    -.4733877

.2887758

       x3   |  -.0031832   .2545593    -0.01   0.990    -.5021103

.4957439

       x4   |   .2289895   .2351569     0.97   0.330    -.2319095

.6898885

       x5   |   .2321194   .1386073     1.67   0.094    -.0395461

.5037848

       x6   |   -.052605   .0149259    -3.52   0.000    -.0818592

-.0233508

       x7   |   -.044577   .0101308    -4.40   0.000    -.0644329

-.024721

       x8   |   .9073838   .5661389     1.60   0.109    -.2022281

2.016996

       x9   |   3.787379   1.812603     2.09   0.037     .2347426

7.340016

       x10  |    .018216   .3453795     0.05   0.958    -.6587154

.6951475

       x11  |   .1276374   .3505633     0.36   0.716    -.5594541

.8147288

      _cons |  -34.79901   15.58228    -2.23   0.026    -65.33972

-4.258305

------------------------------------------------------------------------------

As you can see there are a few independent variables that are not

significant at .1 level.

The following is the result after mfx.

. mfx

Marginal effects after poisson

     y  = Predicted number of events (predict)

        =  .51368381

------------------------------------------------------------------------------

variable |      dy/dx    Std. Err.     z    P>|z|  [    95% C.I.   ]      X

---------+--------------------------------------------------------------------

   x1   |   .0458839      .06442    0.71   0.476  -.080375  .172142

1.52727

   x2   |  -.0474161      .10046   -0.47   0.637  -.244321  .149489

.044021

   x3   |  -.0016352      .13076   -0.01   0.990  -.257914  .254643

2.98182

   x4   |   .1176282      .12433    0.95   0.344  -.126053  .361309

2.30909

   x5   |    .119236      .06652    1.79   0.073  -.011137  .249609

1.67273

   x6   |  -.0270223      .00895   -3.02   0.003  -.044562 -.009483

51.9964

   x7   |  -.0228985      .00574   -3.99   0.000  -.034146 -.011651

59.5873

   x8   |   .4661084      .25856    1.80   0.071  -.040655  .972871

5.27121

   x9   |   1.945515      .91814    2.12   0.034   .146001  3.74503

8.88313

   x10 *|    .009359      .17759    0.05   0.958  -.338702   .35742

.490909

   x11 *|   .0666069      .18605    0.36   0.720  -.298046   .43126

.381818

------------------------------------------------------------------------------

(*) dy/dx is for discrete change of dummy variable from 0 to 1

The statistical significance did not change although marginal effects at

mean and their corresponding standard errors changed.

The following is the result of average partial effects using margeff

. margeff

Average partial effects after poisson

     y  = E(y) (expected number of counts)

------------------------------------------------------------------------------

   variable |      Coef.   Std. Err.      z    P>|z|     [95% Conf.

Interval]

-------------+----------------------------------------------------------------

     x1     |   .0926945   .0066891    13.86   0.000     .0795841

.1058048

     x2     |  -.0956625   .0103587    -9.23   0.000    -.1159653

-.0753598

     x3     |   -.003299   .0135354    -0.24   0.807    -.0298279

.02323

     x4     |   .2393958   .0129356    18.51   0.000     .2140425

.2647492

     x5     |   .2427261   .0077471    31.33   0.000     .2275421

.25791

     x6     |  -.0545182   .0008759   -62.25   0.000    -.0562348

-.0528015

     x7     |  -.0461981   .0006235   -74.10   0.000    -.0474201

-.0449761

     x8     |   .9403796   .0307734    30.56   0.000     .8800649

1.000694

     x9     |   3.925102    .100002    39.25   0.000     3.729102

4.121102

     x10    |   .0190514   .0187024     1.02   0.308    -.0176047

.0557075

     x11    |   .1410915   .0211994     6.66   0.000     .0995414

.1826415

------------------------------------------------------------------------------

As you can see, those insignificant variables turned out to be very

statistically significant.

Why does this happen?

Thanks

SR

*
*   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/


© Copyright 1996–2018 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   Site index