margin effects for continuous variables

[Pilar Giraldez Puig]

Hello,

I need to understand how to interpret the margins effects when my independent variable is continuous in an ordered logit model.
When I run:

xtologit ratingscore zfboard i.naudit zboardindep i.loss i.financialutility zinterestcoverage zTangibility i.Duality zleverage zsize zroaper zMtB i.year, nolog or

margins, dydx (zfboard) vsquish

my result was:

Average marginal effects Number of obs = 5,816
Model VCE : OIM

dy/dx w.r.t. : zfboard
1._predict : Pr(1.ratingscore), predict(pr outcome(1))
2._predict : Pr(2.ratingscore), predict(pr outcome(2))
3._predict : Pr(3.ratingscore), predict(pr outcome(3))
4._predict : Pr(4.ratingscore), predict(pr outcome(4))
5._predict : Pr(5.ratingscore), predict(pr outcome(5))
6._predict : Pr(6.ratingscore), predict(pr outcome(6))
7._predict : Pr(7.ratingscore), predict(pr outcome(7))

------------------------------------------------------------------------------
| Delta-method
| dy/dx Std. Err. z P>|z| [95% Conf. Interval]
-------------+----------------------------------------------------------------
zfboard |
_predict |
1 | -.0190579 .0027301 -6.98 0.000 -.0244089 -.013707
2 | -.0082591 .001247 -6.62 0.000 -.0107031 -.0058151
3 | -.0055766 .001037 -5.38 0.000 -.0076091 -.003544
4 | .0184952 .0027133 6.82 0.000 .0131771 .0238133
5 | .0130614 .001935 6.75 0.000 .0092688 .016854
6 | .0011374 .0002618 4.34 0.000 .0006243 .0016505
7 | .0001997 .0000736 2.71 0.007 .0000553 .000344
------------------------------------------------------------------------------

.
end of do-file

As zfboard is a continuous variable, I guess that, for instance, for the first level on the dependent variable, one percent increase in zfboard diminish the probability of being in that level a -.0190579 per cent.
Please, could anyone tell me if I am right?

[Marcos Almeida]

Welcome to the Stata Forum / Statalist.

If I understood your query correctly, in #1 we have the mean predicted values of zfboard for each level of ratingscore.

[Pilar Giraldez Puig]

Thank you, Marcos. I read the paper from Williams, (2012) but it is only for categorical variables, so I hardly understood the meaning in the case of continuos independent variables. I also tried to consult Cameron and Trivedi (2010), but it is not available yet on my library.

[Marcos Almeida]

Maybe this presentation will be helpful to you, particularly the pages 11 and 12.

[Pilar Giraldez Puig]

Guau!!
Thank you so much!