I am trying to make some observations from a data set of independent and dependent variables. Using its S and AD Test statistic, the set of natural logarithms of the dependent variables has a P-value of 0.7586. I believe this means I can assume the dependent variables are distributed normally.
A simple regression analysis yields:
r^2 0.6612
Significance F 0.000
P-value 0.000
Number of observations 76
Average of X (independent variable) 15.86
Sample intercept 2.16
Sample slope (0.12)
Sum of residuals^2 16.76
Std deviation of residuals 0.48
SS(X) 2439.72
Sumproduct all Xs 21,563.08
Sum all Xs 1,205.56
All rounded to 2 digits
Assuming X=20, the point estimate is (0.1774)--which is 0.837 when converted from Ln. By using the following equation, I believe I can assume Y will be at least 0.267 99% of the time when X=20:
=EXP(+Ln point estimate for Assumed X-((TINV(1-98%,(Count-2)))*(SQRT(Sum of residuals^2/(Count-2)))*SQRT(1+1/Count+(Assumed X-Average X)^2/(SUMPRODUCT(All Xs, All Xs)-SUM(All Xs)^2/Count))))
Am I on the right track?
A simple regression analysis yields:
r^2 0.6612
Significance F 0.000
P-value 0.000
Number of observations 76
Average of X (independent variable) 15.86
Sample intercept 2.16
Sample slope (0.12)
Sum of residuals^2 16.76
Std deviation of residuals 0.48
SS(X) 2439.72
Sumproduct all Xs 21,563.08
Sum all Xs 1,205.56
All rounded to 2 digits
Assuming X=20, the point estimate is (0.1774)--which is 0.837 when converted from Ln. By using the following equation, I believe I can assume Y will be at least 0.267 99% of the time when X=20:
=EXP(+Ln point estimate for Assumed X-((TINV(1-98%,(Count-2)))*(SQRT(Sum of residuals^2/(Count-2)))*SQRT(1+1/Count+(Assumed X-Average X)^2/(SUMPRODUCT(All Xs, All Xs)-SUM(All Xs)^2/Count))))
Am I on the right track?