Serving the Quantitative Finance Community
Can we calculate t-statistics when objective function is minimizing pricing error?
We can calculate it when the objective function is log-likelihood, I am wondering if we can do so when the objective is pricing error of a nonlinear function, say, RMSE (root mean squared error). The definition of t-value doesn't stop us computing it, however, I notice the parameters' standard error changes when we scale the objective function (and thus the t-value), for example, multiplying RMSE by 100, etc.Thanks a lot.