Finance and economics rarely give us ‘ideal’ data sets to play with. Those who know a little about stats frequently point out, for example, that in finance there are fat tails (t distributed errors), and go on to claim that this invalidates the parametric tests we use to test hypotheses.
For a good reminder of why this is not the case, I recommend this excellent
I’ll have a read of that post. One comment I would make is that criticism of using parametric tests that use normality assumption is that they generally also assume independence of observations (independence of some strength). This is clearly violated by economics data.
Comments are closed.