Error calculating marginal effects with a tobit model

[Max Ortel]

I am looking at the effect of permitting standards, rebates, average income, and electricity price on a counties solar capacity (cap). Below is an abbreviated sample of my data.

* Example generated by -dataex-. To install: ssc install dataex
clear
input float(cap permit rebate lnincome electric) double houses
1452.547 0 47.61906  10.23742  10.88  578.2945736434108
595.21027 1     5500 10.662493 15.142 14524.418604651162
545.2054 1     5500 10.641537 13.932 14359.302325581395
520.4196 1     5500  10.77956 13.932  7180.232558139534
502.8186 1     5500 10.793516 15.142  7281.782945736434
474.9224 1     5500  10.67165 15.142 1148.8372093023256
398.07135 1     5500 10.635086 13.932 1160.4651162790697
306.4947 0 1.00e-05  10.38539  10.99   4659.68992248062
264.1102 1     5500 10.537601 13.824 14199.612403100775
239.5443 0 1.00e-05 11.136908  10.99   315.891472868217
end

I run the following command

tobit cap permit rebate lnincome electric [aweight = houses], ll robust

I then run the following command:

margins, eydx(rebate) atmeans

I receive the following error:

"could not calculate numerical derivatives -- discontinuous region with missing values encountered"

the following command does work however:

margins, dydx(rebate) atmeans

Would love some ideas about what going on here, thanks so much in advance!

[Maarten Buis]

I am not surprised as rebate seems to have an extreme scale (5500 to 0.000001 in just your example). You will probably need to change that to something more reasonable, but what reasonable is depends on the substantive meaning of that variable, so that is up to you.