Regression Predictions Forum

[jpSartre]

I ran a regression on my PTs over time, and the trend predicts a 172 come test day. Anyone interested enough to compare?

[voice of reason]

What is the standard error of the estimate?

[jpSartre]

2.6. It's been a while since econ stats... does that mean 172 +/- 2.6, are there degrees of freedom involved or am I confusing topics entirely?

[mountaintime]

my 172 turned into a 168 on test day

[voice of reason]

2.6. It's been a while since econ stats... does that mean 172 +/- 2.6, are there degrees of freedom involved or am I confusing topics entirely?

Once you have a standard error calculated you don't have to think about degrees of freedom anymore.

It means the 95% confidence interval for the predicted score on your next practice test is 167 to 177 (172 +/- 2.6*1.96). In other words, if 100 people had the same practice record as you, then 95 of them would get a score in the range of 167 to 177 on the next test.

In practice the band may be narrower than that, because the test is designed to be a reliable measure of latent ability. To be sure I'd have to think more carefully about the statistical theory than I want to on a Friday night.

[jpSartre]

2.6. It's been a while since econ stats... does that mean 172 +/- 2.6, are there degrees of freedom involved or am I confusing topics entirely?

Once you have a standard error calculated you don't have to think about degrees of freedom anymore.

It means the 95% confidence interval for the predicted score on your next practice test is 167 to 177 (172 +/- 2.6*1.96). In other words, if 100 people had the same practice record as you, then 95 of them would get a score in the range of 167 to 177 on the next test.

In practice the band may be narrower than that, because the test is designed to be a reliable measure of latent ability. To be sure I'd have to think more carefully about the statistical theory than I want to on a Friday night.

Haha, understandable. I appreciate the explanation.

I've got 3 tests left so well see if I can't bump that confidence interval up.