07-19-2017 03:27 PM
I am building a program to interpolate missing bond yield values between years. There is some sensitivities to accuracy with this type of financial data and so I need to be sure that at the very least, I am performing a cubic spline interpolation - polynomial interpolation of order 3 fitted with splines.
In using PROC EXPAND, will interpolating fill out any blanks that I ask it too, versus creating arrays, which require the same pattern of missing values? Can someone clarify this difference?
Would someone recommend that PROC EXPAND is the best way to interpolate this?
Lastly, even in reading the SAS.HELP for PROC EXPAND, I am not sure I understand the set-up for the frequency of the data that needs to be interpolated. For me, it is evenly-spaced yearly intervals.
07-19-2017 06:14 PM
I've used Proc Expand to do interpolation with cubic splining so I would have thought it was the way to go for you. There are lots of different parameters and options though so you'll have to decide which of them to use and yes, it will definitely fill in missing rows and I think to get end of year intervals you'd specify ALIGN=END and TO=YEAR in the Proc Expand call.
07-20-2017 11:26 AM
Can I ask how you decided that cubic spline interpolation was better for your needs than a linear interpolation??
07-20-2017 06:09 PM
To be honest that wasn't my decision, as a jobbing programmer I just had to implement it. We were using it to convert annual series to quarterly series to benchmark against other quarterly series.