Reply
Thu 9 Aug, 2012 01:25 pm
I have a set of 75 independent and dependent variables. The independent variables are collected each January 1 and the dependent variables represent a value measured 37 years after the corresponding X value is collected. There is a high correlation between X and Y but there is also autocorrelation. My goal is to calculate the confidence limits of a predicted Y from a particular X.
I can use various ARIMA methods to correct for the autocorrelation but I believe ARIMA only allows me to predict Y for the next X in the time series. Since the Y resulting from last year's X will not be known for 36 years, ARIMA will not help me unless I use the 37-year actual change in X to predict the 37-year change in Y.
I have tried to transform the X and Y variables to eliminate the autocorrelation but Durbin-Watson test still shows autocorrelation.
Any way I can calculate the confidence limits of a predicted Y from a particular X?
@cisaak,
=So lets see of I understand--I collect an X each year for 75 years and measure a Y corresponding to each sequential X, 37 years into the experiment.
my data sheet looks like this with a 37 year dead space
Y X Y
1 x0
2 x1
3 x2
. .
. .
. .
37 x37 y1
38 x38 y2
39 x39 y3
. .
. .
. .
75 x75 y38
76 y39
77 y40
. .
. .
. .
109 y73
110 y74
111 y75
Rap
@raprap,
You are correct. My theory is y1 is dependent on x1, y2 is dependent on x2, ... y37 is dependent on x37.