Who's panicking?

[SDSUAlum08]

Re: Who's panicking?

Actually not. Standard error of the mean is not given. A quick calculation for the Summit League teams results in a standard error of 0.0175. Given that a 95% confidence interval about the mean would then be about +/-0.035. So the 95% mean confidence interval for the Summit RPI is 0.517 to 0.447. A quick look at conference RPIs and one can conclude no statistically significant difference in mean RPI between the Summit League and the WAC at No. 123 and the Big South at No. 27.


Hence the problem with using RPI for comparisons of conferences or for that matter teams.

My head hurts. I understand statistics, but I hated that class in college and in grad school....yuck!

Although, nicely done! I'd crunch the numbers to confirm...but...um....yeah...

[zooropa]

Re: Who's panicking?

Actually not. Standard error of the mean is not given. A quick calculation for the Summit League teams results in a standard error of 0.0175. Given that a 95% confidence interval about the mean would then be about +/-0.035. So the 95% mean confidence interval for the Summit RPI is 0.517 to 0.447. A quick look at conference RPIs and one can conclude no statistically significant difference in mean RPI between the Summit League and the WAC at No. 123 and the Big South at No. 27.


Hence the problem with using RPI for comparisons of conferences or for that matter teams.

But, and feel free to enlighten me on this, is it proper to use a 'confidence interval' about a mean when the entire population has been sampled? That is, the mean RPI for the Summit League is not calculated using a representative sample of SL teams, it's calculated using the entire population.... Therefore, can't it be assumed that there is no need for a confidence interval about the mean?

And, again, feel free to educate me on this one.

I was basing the assertion of statistical significance based on the fact that we know the means for these conferences with certainty, and that in that 12-27 range, a very small change in the mean results in a significant change in ordering, although in the strictest sense, that's probably not the same as 'statistically significant'.

[filbert]

Re: Who's panicking?

But, and feel free to enlighten me on this, is it proper to use a 'confidence interval' about a mean when the entire population has been sampled? That is, the mean RPI for the Summit League is not calculated using a representative sample of SL teams, it's calculated using the entire population.... Therefore, can't it be assumed that there is no need for a confidence interval about the mean?

And, again, feel free to educate me on this one.

I was basing the assertion of statistical significance based on the fact that we know the means for these conferences with certainty, and that in that 12-27 range, a very small change in the mean results in a significant change in ordering, although in the strictest sense, that's probably not the same as 'statistically significant'.

And boom goes the stat-geek dynamite.

[bigticket1]

Re: Who's panicking?

And boom goes the stat-geek dynamite.

I know I'll sleep better tonight because of knowing all this.

[jackrabit1]

Re: Who's panicking?

Oh yeah thats my favorite school.

I think Nidaros found his IPFW!

[Nidaros]

Re: Who's panicking?

I think Nidaros found his IPFW!

Yeah I love all the Arkansas schools. Especially the Razorbacks. I though I thought I would wound the pride of that individual who hate us northerners. The nerve to put down our beloved SF. UALR wow, what a power house.

[Prairiehaas]

Re: Who's panicking?

But, and feel free to enlighten me on this, is it proper to use a 'confidence interval' about a mean when the entire population has been sampled? That is, the mean RPI for the Summit League is not calculated using a representative sample of SL teams, it's calculated using the entire population.... Therefore, can't it be assumed that there is no need for a confidence interval about the mean?

And, again, feel free to educate me on this one.

I was basing the assertion of statistical significance based on the fact that we know the means for these conferences with certainty, and that in that 12-27 range, a very small change in the mean results in a significant change in ordering, although in the strictest sense, that's probably not the same as 'statistically significant'.

Sorry, been a bit busy. The confidence interval arises from the error in measurement. Think of the RPI calculation as you would a ruler. There is some error in that measurment. Therefore the statisitics cannot be precise.

OK, nerd mode off.

[Nidaros]

Re: Who's panicking?

Sorry, been a bit busy. The confidence interval arises from the error in measurement. Think of the RPI calculation as you would a ruler. There is some error in that measurment. Therefore the statisitics cannot be precise.

OK, nerd mode off.

No they sure are not, and this coming from some one who struggled to get a C in a major course at SDSU, years ago. I do understand zoo statement about RPI being sampled at 100 per cent of the population. There are what roughly 350 D1 schools and it different than auditing something like insurance claims for State Farm which would include a popluation of thousands if not millions of claims. In auditing the 95 per cent confidence level is a big part of of the statistcal methods in accepting a sample, and I see why now.

[filbert]

Re: Who's panicking?

The problem with applying the concept of confidence interval to the RPI is that the RPI is not a probabilistic statistic. That is, there are no probabilities involved in the calculation of the RPI (or; all probabilities are 100%; and you can think of calculating confidence intervals by in essence multiplying the probabilities together. The RPI is simply a somewhat clever method of confounding some degree of strength-of-schedule with a team's own won-loss percentage. The probability that SDSU's 2011-12 W-L record was 27-8 is 100%. Similiarly, the probability of the sums of all of SDSU's opponents' and opponent's opponents' W-L records is also 100%.


You can multiply 1 by 1 as many times as you want, the result is always 1. There's no room in the RPI for any confidence interval whatsoever. It is what it is.

It is a combined measure of a team's W-L record with the W-L records of it's opponents and the opponent's opponents. When considered using that definition, it is what it is. But when it's interpreted as some kind of measure of relative team strength, you depart from the strict mathematics of the RPI and enter epistemology. Or perhaps metaphysics or aesthetics. One of those sub-topics of philosophy, anyway.

After posting, I realized I failed to close my aside-parenthesis. Here: )

[Nidaros]

Re: Who's panicking?

Right on. Thanks you helped this guy who got a C in statistics even though my offical major was commerical economics.

[LakeJack]

Re: Who's panicking?

This may have turned into the most boring thread ever. But you stats geeks need an outlet too I guess.

[zooropa]

Re: Who's panicking?

the error in measurement. Think of the RPI calculation as you would a ruler.

I understand, except nothing is being measured. The RPI is established by definition, not by measurement.

However, I'm pretty sure I shouldn't've said that the difference was statistically significant.

[filbert]

Re: Who's panicking?

This may have turned into the most boring thread ever. But you stats geeks need an outlet too I guess.

"Math is hard!" - infamous "Talking Barbie" quote.


Maybe we should turn this into one of those old Schoolhouse Rock things . . .

[LakeJack]

Re: Who's panicking?

I'm just a bill . . .

http://www.youtube.com/watch?v=H-eYBZFEzf8

[RabbitObsessed]

Re: Who's panicking?

I was always partial to Conjunction Junction, but that may be my English major side.