serial correlation, heteroskedacity, et al

I am consistently mixing up serial correlation, heteroskedacity, multicolinearity and the tests to check for them and the actions to take when discovered…am i alone? Can anyone suggest a simple way (mnemonic or rhyme) to remember them. Am seriously snapping out over the volume of testable details in Level II, not to mention I took a pre-review practice exam and scored 53%–not terrible uplifting. Thanks.

I recognize that this is lame, but I remember that Multicollinearity results in too many type II errors because “multi” means more than 1, i.e. 2, so not type 1, but type 2…

its not that lame as i am pretty sure i said type one on a practice the other night.

Positive serial correlation has the similar effect as hetero***, the result will be se too small, t-test too large, and more TYPE I error.

All of this comes down the what happens to the standard errors. Coeff/s.e.=t-stat If s.e. are too small then t-stat too big (less discounting in the denom) and Type I error Heteroskedacity Pos autocorrelation If s.e are too big then t-stat too small (more discounting in the denom) and Type II error Multicollinearity neg autocorrelation

what about all the dickey fuller, durbin watson, breusch pagen blah blah blah…short of pure memorization, how do you keep it straight?

Heteroskedasticity is the first issue mentioned in the text. The first thing that ballplayers do is take BP. That’s how I get that sorted out in my head. If I didn’t think of heteroskedasticity first, I don’t know how i’d get them in my head.

and dickey… unit… get it? test for a unit root.

SanFranMatt Wrote: ------------------------------------------------------- > Heteroskedasticity is the first issue mentioned in > the text. The first thing that ballplayers do is > take BP. That’s how I get that sorted out in my > head. > thats fantastic.

> and dickey… unit… get it? test for a unit root. brilliant, banny.

lol

I’m also having trouble keeping them straight… but these pointers will definitely help me out. thanks! :slight_smile:

I am not proud of this one but Serial correlation is the SERIAL killer Sherlock Holmes, detective, had a nice pal named Watson with who he was chasing serial killers So Durbin WATSON helps to track and correct SERIAL correlation

nicolargol Wrote: ------------------------------------------------------- > I am not proud of this one but Serial correlation > is the SERIAL killer > > Sherlock Holmes, detective, had a nice pal named > Watson with who he was chasing serial killers > > So Durbin WATSON helps to track and correct SERIAL > correlation Damn, that’s a good one. I use this: Durbin Watson sounds like an old kind of car. LIke “I drive an old 1954 Durbin Watson”. A car is an auto. Auto-correlation is the same thing as serial correlation.

pardon my ignorace, but what does BP stand for? bullpen? base…?

Breusch-Pagan test.

This one’s a little weak, but it works for me. Hetero/homo skedasticity Being a hetero myself, when I see the word hetero, what do I think of? Sex. What does sex do? Gets my heart pumping. What does that do? Affects my Blood Pressure. = BP

haha… thats awesome!

I meant in response to SanFranMatt… “The first thing that ballplayers do is take BP. That’s how I get that sorted out in my head.”

in baseball, BP means Batting Practice. It’s what they do before games as a warm-up.