I was wondering where we get the "t(N-2)(1-alpha) value for the CI? On our slides, the example gives 1.97 for this value...so are we to use this every time we need a 95% CI...what is the value then if we want a 99% CI?
t(N-2) is from t-test. You use t-test when you don't know the population variance and you have a small sample size (lecture 8 slide 5).
(1-alpha) = (1-.05) = .95 <--the alpha here is referring to significance level. You use this when you are calculating 95% CI. 95% CI means that 95 out of 100 times that you predict a score in Y in the population will fall into the range of score that you predict from a regression model.
So, if you want 99% CI, alpha will be .01. So, (1-alpha) = 1-.01 = .99.
Oh, I'm sorry that I answered half of your question earlier. David is right, you can to find the t-value by looking up from a table. But SPSS can do that work for you too. 1.97 from the example in the lecture ppt is actually from t(N-2)*(1-.05)*(SE*standardized B). So, the value change whenever you have a different t score, N, alpha level, standard error and standardized beta. So, the values does not have to be 1.97 all the time, it depends on your regression model. You can find the formula for CI on lecture 8 slide 9.
If i understand correctly, you're wondering where the t comes from...that's the t score that has its own formula. This is why it will change between data sets. It has nothing to do with whether or not you want 99 or 95% confidence. That is adjusted using the alpha level (as Grace displayed above). But we mostly use 95 so you shouldn't have to worry too much about that.
5 comments:
t(N-2) is from t-test. You use t-test when you don't know the population variance and you have a small sample size (lecture 8 slide 5).
(1-alpha) = (1-.05) = .95 <--the alpha here is referring to significance level. You use this when you are calculating 95% CI. 95% CI means that 95 out of 100 times that you predict a score in Y in the population will fall into the range of score that you predict from a regression model.
So, if you want 99% CI, alpha will be .01. So, (1-alpha) = 1-.01 = .99.
Hope this helps!
Well, I think you would have to look the t value up in a table. The value depends on the alpha but also the degrees of freedom.
http://changingminds.org/explanations/research/analysis/t-test_table.htm
Oh, I'm sorry that I answered half of your question earlier. David is right, you can to find the t-value by looking up from a table. But SPSS can do that work for you too. 1.97 from the example in the lecture ppt is actually from t(N-2)*(1-.05)*(SE*standardized B). So, the value change whenever you have a different t score, N, alpha level, standard error and standardized beta. So, the values does not have to be 1.97 all the time, it depends on your regression model. You can find the formula for CI on lecture 8 slide 9.
Thanks, I've memorized the formula from the slide, but didn't know what the t value would be for 99% for alpha instead of 95%.
If i understand correctly, you're wondering where the t comes from...that's the t score that has its own formula. This is why it will change between data sets.
It has nothing to do with whether or not you want 99 or 95% confidence. That is adjusted using the alpha level (as Grace displayed above). But we mostly use 95 so you shouldn't have to worry too much about that.
I hope i understood the confusion well enough!
Post a Comment