Closed hcwi closed 6 years ago
hi @hcwi would you be able to provide more background information? are you asking to create subtypes of 'ndf' , one for the 'model' and the other for the 'error' (as discussed in https://math.stackexchange.com/questions/617735/multiple-regression-degrees-of-freedom-f-test) ?
Despite closing the issue, let me summarise the discussion:
F statistic is by definition a ratio of two mean squares, each with its df number. The multiple regression is a special case of this, where "numerator" is the mean square for the whole model, and "denominator" is the means square for error. In general, F is for testing general hypotheses, e.g. corresponding to a "term", not necessarily being the whole model.
As to computation of both df numbers - they can be exact or approximated by the approaches already included in STATO.
And thanks for addressing the request!
Hi, I am missing from STATO the terms 'numerator degrees of freedom' and 'denominator degrees of freedom' to quickly distinguish between 'number of degrees of freedom' (STATO_0000069) for F-distribution and F-tests. Unless I have overlooked some other way to do this, it might be useful to add them.