I have a statistics test tonight

[Roger_the_Shrubber]

Yes, I meant minimum. And 90% confidence is fine.

[ggil]

<BLOCKQUOTE><font size=2 face="Verdana, Helvetica, sans-serif">quote:</font><table border="0" width="90%" bgcolor="#333333" cellspacing="1" cellpadding="0"><tr><td width="100%"><table border="0" width="100%" cellspacing="0" cellpadding="2" bgcolor="#FF9900"><tr><td width="100%" bgcolor="#DDDDDD"><font size=2 face="Verdana, Helvetica, sans-serif">Originally posted by Python:
I prefer the left hand variable so I can work the mouse with my right hand.

Trust me, I've done extensive research in this area.</font></td></tr></table></td></tr></table></BLOCKQUOTE>

...and from a mile away no less.

Rack the obvious

oh the humanity

[Adept]

<BLOCKQUOTE><font size=2 face="Verdana, Helvetica, sans-serif">quote:</font><table border="0" width="90%" bgcolor="#333333" cellspacing="1" cellpadding="0"><tr><td width="100%"><table border="0" width="100%" cellspacing="0" cellpadding="2" bgcolor="#FF9900"><tr><td width="100%" bgcolor="#DDDDDD"><font size=2 face="Verdana, Helvetica, sans-serif">Originally posted by Dr Bob:
[QB]OK, Smackie, at what sample size would you feel confident invoking the Central Limit Theorem?


And, Adept, simply having more than than one explanatory variable doesn't necessarily mean you run to the shelter of the F-stat V the t-stat, it is inherently a function of the hypothesis you are in the process of testing.


For example, I may want to understand the impact of taxation of GNP, but I have to include other right hand side variables to avoid excluded variable bias....but my focus will be on the coefficient and t-stat for my taxation variable, where I won't care as much about an F-stat or the R-squared and shit.


Edit: AND, the F and t-tests would basically converge with only one right hand side variable....(F is actually t squared with one RHS)...</font></td></tr></table></td></tr></table></BLOCKQUOTE>

After 30 observations you use a normal distribution over the student-t.

Your attempt at proving me wrong was pretty lame. If you just wanted to know the effect on taxation on GNP than just regress the two, or algebracially remove taxation from the other variables. Why bother throwing all of that other shit in - it just comprimises the regression.

So in reality a t score is only used when comparing one RHS variable, and a F-Stat is used for more than one RHS variable.

Also, Dr. Bob, who the hell cares about GNP anymore? what is the marginal propensity to import in the US? If you regressed imports you would find that the value would be significant at both 5% and 1% levels!

[Dr Bob]

<BLOCKQUOTE><font size=2 face="Verdana, Helvetica, sans-serif">quote:</font><table border="0" width="90%" bgcolor="#333333" cellspacing="1" cellpadding="0"><tr><td width="100%"><table border="0" width="100%" cellspacing="0" cellpadding="2" bgcolor="#FF9900"><tr><td width="100%" bgcolor="#DDDDDD"><font size=2 face="Verdana, Helvetica, sans-serif">Originally posted by Adept:


After 30 observations you use a normal distribution over the student-t.

Your attempt at proving me wrong was pretty lame. If you just wanted to know the effect on taxation on GNP than just regress the two, or algebracially remove taxation from the other variables. Why bother throwing all of that other shit in - it just comprimises the regression.

So in reality a t score is only used when comparing one RHS variable, and a F-Stat is used for more than one RHS variable.

Also, Dr. Bob, who the hell cares about GNP anymore? what is the marginal propensity to import in the US? If you regressed imports you would find that the value would be significant at both 5% and 1% levels!</font></td></tr></table></td></tr></table></BLOCKQUOTE>

It was a hypothetical example, jeez, we've gotten a bit touchy, huh?

Lame? Listen, I merely pointed out that the t-test and the F-test test TWO COMPLETELY DIFFERENT HYPOTHESES.

AND, as for your argument that you merely need to include a single variable on the RHS, why don't you take the time to go back and check the assumptions of the multiple regression model (specifically, the one that says the estimates are BLUE if and ONLY if the model is correctly specified....hence there are very few good models with one RHS variable).

It is stupid and wrong to think that a t-test and F-test are somehow mutually exclusive (especially since I just told you that they are so absolutely related to each other in the one RHS variable case). I could right now go find one thousand very good scientific research articles that use a t-test to test a specific hypothesis within the structure of a model with multiple independent variables.

Oh, by the way, you do NOT switch to a z-test at n=30. The t-test is used because we don't know the true population variance so we use sigma-hat instead (and thus the t-test with n-2 degrees of freedom). The t-statistic and z-statistic don't converge until you have a much larger N (although the two are sufficiently close that you wouldn't fuck up too bad by flipflopping the two).


Let me guess, you've had a class or two in this stuff as an undergrad. Lemme know if you want to continue to discuss such matters.

[Python]

I wish I was ambidextrous.

[Adept]

Way to mask your insecurity with 'whitty' (is it witty?) jokes.

[Dr Bob]

Who wants to bet me $50 that Py looked up "ambidextrous" on some online dictionary before posting that....




OR he typed it in a Word document to see if Bill Gates's Red Underscore of Humility appeared....

[Dr Bob]

<BLOCKQUOTE><font size=2 face="Verdana, Helvetica, sans-serif">quote:</font><table border="0" width="90%" bgcolor="#333333" cellspacing="1" cellpadding="0"><tr><td width="100%"><table border="0" width="100%" cellspacing="0" cellpadding="2" bgcolor="#FF9900"><tr><td width="100%" bgcolor="#DDDDDD"><font size=2 face="Verdana, Helvetica, sans-serif">Originally posted by Adept:
Way to mask your insecurity with 'whitty' (is it witty?) jokes.</font></td></tr></table></td></tr></table></BLOCKQUOTE>


Way to change the subject to divert attention....

[Python]

What I want to know is since when did Merriam Webster's Online Dictionary require a username and password? I had to walk all the way down the hall to borrow a dictionary.

Rack the good doctor.

[Adept]

I didn't mean to say that they were mutally exculsive, just that in the event that you are doing a multiple variable regression you may as well use a F-Stat since it gives better information (i.e. can include R^2 and other shit).

Its like sailing across the atlantic or flying. You will get there, but it is probably a better idea to fly. Very good analogy Adept. Becuase some people may still want to sail dispite additional cost and lost time for some unknown reason, and they are still welcome to do that.

So what you said about switching from t to z is exactly what i said? After 30 d.f. the standard normal is 'close enough' to the student-t that you can flip flop between the two without any significant negative effects.

Classes is the wrong word. A couple courses as an undergrad - yeah - whats your point - you really haven't asserted any sort of superior knowledge because i am a mere undergrad!