or Connect
Styleforum › Forums › General › General Chat › Anyone a 'stats geek'?
New Posts  All Forums:Forum Nav:

Anyone a 'stats geek'? - Page 2

post #16 of 24
Quote:
Originally Posted by patrickBOOTH View Post

This is pretty on point, but there are still many more things to consider such as getting the data specification, for example a lot of these kind of variables have the common issue of being tied to time. Over time these types of variables just go up so if you regress all of them together you are going to have insane multicolliniarity and inflated R^2 regardless of using adjusted or not. You have to break the trends by differencing, % change, log and so on. Even with that you have to test your VAR's and so on. It is a much more intense process than I feel you can get through an online forum if you want to do it right.
Also this last paragraph is important a model is not a forecast. A model is an explanation of what happened at a point in time, this may or may not be applicable to the future.

I don't think he has time series data, but if he does you are correct. There should, as always, be tests of multicollinearity before running the model. Also, depending on what the variables look like, he probably needs to break things like "demographics" into several dichotomous variables since "demographics" is probably a nominal non-ordered variable. Finally, like I said above, fixed effects would probably be a serious consideration.

In short, I think this thread shows that there is a lot more to statistics than just throwing variables into a model. Good statistics can be very helpful and tell us some interesting things about the world, but I really believe that I'd rather have no statistics than bad statistics.
post #17 of 24
Quote:
Originally Posted by John152 View Post

I don't think he has time series data, but if he does you are correct. There should, as always, be tests of multicollinearity before running the model. Also, depending on what the variables look like, he probably needs to break things like "demographics" into several dichotomous variables since "demographics" is probably a nominal non-ordered variable. Finally, like I said above, fixed effects would probably be a serious consideration.
In short, I think this thread shows that there is a lot more to statistics than just throwing variables into a model. Good statistics can be very helpful and tell us some interesting things about the world, but I really believe that I'd rather have no statistics than bad statistics.

This is all true, a log-log transformation would be good for cross sectional data, however you are correct sound theory always makes for a better model.
post #18 of 24
Holy cow, hate to interrupt but what did everyone major in/current profession? I'm a marketing student and I got pretty interested in stats myself smile.gif
post #19 of 24
Finance, economics, psychology.

I still don't know what is wrong with me .
post #20 of 24
Haha, I'm marketing, entr., and economics - though entr as a major is total BS in my opinion.
post #21 of 24
What the hell is entr?
post #22 of 24
Sorry for my laziness: "entrepreneurship"
post #23 of 24
Ah. Is the first lesson on how to start with a lot of money?
post #24 of 24
Haha it's a pretty BS curriculum - I think there is literally a class that is about using money from your parents. The rich foreign kids were pretty pumped.
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: General Chat
Styleforum › Forums › General › General Chat › Anyone a 'stats geek'?