quant: prediction interval

[mp3bu]

If we’re given an ANOVA table, and it asks for a prediction interval, should we use:
mean +/- (given standard error)(t stat)
or should we replace the given standard error with sqrt((standard error^2)((1 + (1/n) + ((x - mean)^2)/((n-1)(variance))))
I don’t understand when to use one or the other.

 

[Iginla2010]

When n is too large you can use the given error value. If not, then unfortunately you’ll have to calculate the satndard error of the estimate using the formula you mentioned above.

 

[mp3bu]

Oh OK. Got it. Is there a magical “n” figure? 30? 100?

 

[cpk123]

difference is prediction interval for a variable in the regression (which is the first equation)
and 2nd equation is prediction interval for a forecast made using the regression.