Interpretation of variable in a probit model

[naveed ahmed]

Dear Statalisters,

I am struggling interpreting the coefficient of a variable which is expressed as a proportion in a probit model. As it currently stands , I am interpreting the average partial effects of this variable (0.14) as a one unit increase in share of debt raises the probability of innovation by 14 percentage points.



However, if I multiply this variable by 100 and turn it into a percentage, the average average partial effect is (.0014556 ), does it mean that a 1 percentage point change in the percentage of debt increases the probability of innovation by 0.0014556.



Questions : a) are my interpretations correct

b) Instead of saying a one percentage change increases can I write in terms of a 10 percentage point increase and if so what should I do .


Regards,
naveed

[Carlo Lazzaro]

Naveed:
see the interesting debate on a similar issue at: https://www.statalist.org/forums/for...evel-log-model

[Phil Bromiley]

In some cases, it is easier to use margins to generate changes in predicted probabilities for specific changes in the iv's. This circumvents the non-linear problem somewhat. However, in such non-linear models, the influence of a given variable depends on the values of the other variables. Often this is not a problem (doesn't make a big difference), or one simply talks about the influence of x holding all other variables at their means (the default in margins).

[naveed ahmed]

Dear Phil,

I ended up using margins to obtain changes in predicted probabilities. Really grateful for your reply.

Regards,
naveed