Calculate derivative of partial effect in probit with respect to the parameter

283 Views Asked by At

In the standard probit regression we have that $ Pr(y>0| X) = \Phi(\frac{X\beta}{\sigma})$.

With $\Phi$ Cumulative Density Function of the Normal Distribution.

The marginal effect of the covariate $ x_j$ is:

$\phi(\frac{X\beta}{\sigma}) \beta_j$

Where $X$ is the nxk vector of covariates, and $\beta$ is the kx1 vector of coefficients and $\phi$ Probability Density Function of the Normal Distribution.

Now I want to calculate the derivative of this marginal effect with respect to the parameter $\beta_j$, namely

$\frac{\partial \phi(\frac{X\beta}{\sigma}) \beta_j}{\partial \beta_j }$

Using the Chain rule I have:

$\frac{\partial \phi(\frac{X\beta}{\sigma}) \beta_j}{\partial \beta_j } = \phi(\frac{X\beta}{\sigma}) +\frac{\partial \phi(\frac{X\beta}{\sigma})}{\partial \beta_j } \beta_j $

How do I compute $\frac{\partial \phi(\frac{X\beta}{\sigma})}{\partial \beta_j } $ ?