I have been trying to maximize a function that involves an integral of the error function (via the CDF of the normal distribution) with respect to a parameter that appears in the limit of the integration range. When I compute that integral using "NIntegrate" I obtain a solution with "NIntegrate::inumr" warnings. When I compute the using "Integrate" I get no warnings. Both methods produce the required result. I would like to understand why I get the warnings when computing numerically and not symbolically. Can anyone help me this?
P.S. Mathematica file is attached.
Attachment: MyQuestion.nb, URL: ,