can someone explain in a simple sentence what it means to “Specify a regression model a priori” to test on a given data set?
“specify a regression model a priori” = “specify a regression model ‘before’ looking at, taking, or knowing about”
about what? the risk factors? the undistilled data?
would depend on the nature of the data. and the predictive goal of the regression
Basically it means to create the model before you look at the data, rather than look at the data and decide what model fits it best. So you’d need to have some kind of theory or idea about what the model should look like, based on your general knowledge of markets and economics, and perhaps the industry involved. A priori basically means “coming first” as in “before you look at anything.”
why would someone do an a priori regression?
pacmandefense Wrote: ------------------------------------------------------- > why would someone do an a priori regression? to reduce impact of curve-fitting
Depends on the context. 1) It avoids data mining and reduces the chance that you create some model that performs dismally out-of-sample. If the a-priori model is bad, it will perform badly in sample and you probably won’t use it out-of-sample. Too much data mining, and you can get something that looks good but then does badly when you use real money with it. 2) It may just be a test of one’s grasp of finance theory and modeling skills. In real life, there is almost always some adjustment between the theory and the data. The trick is to try to keep from overfitting the data and lapse into data mining, which is a very seductive call. If you are able to think about models a priori, you increase the probability that you actually have an insight that your model can work with.