even if it's a nonlinear model, you can still use "test" after estimation.

testnl is for testing linear or nonlinear hypotheses about the estimated parameters. eg,

. sysuse auto
. generate weight2 = weight^2
. regress price weight weight2
. . testnl _b[mpg] = 1/_b[weight]

here you use testnl, because 1/_b[weight] is nonlinear for the parameter of "weight"

We cannot use Wald test for the significannce of particular varialbes in a nonlinear system.Likelihood ratio test is the alternative method.

how can we know the result from the Likelihood ratio test?

probit
logit
都是非线性的
你看看这些模型的检验方法就知道怎么做了

1. testnl -- Test nonlinear hypotheses after estimation

    testnl tests (linear or nonlinear) hypotheses about the estimated parameters from the most recently fitted model.
    testnl produces Wald-type tests of smooth nonlinear (or linear) hypotheses about the estimated parameters from the most
    recently fitted model.  The p-values are based on the delta method, an approximation appropriate in large samples.
    testnl can be used with svy estimation results, see [SVY] svy postestimation.
    The format (exp1=exp2=exp3= ... ) for a simultaneous-equality hypothesis is just a convenient shorthand for (exp1=exp2)
    (exp1=exp3), etc.
    testnl may also be used to test linear hypotheses. test is faster if you want to test only linear hypotheses.  testnl
    is the only option for testing linear and nonlinear hypotheses simultaneously.


2. test -- Test linear hypotheses after estimation