I work in Workers Compensation, and I frequently make Paid Loss Triangles, so when I finish the course, I need to be able to use this analysis.  I have two questions about the formatting.

First, the article says "divide by the measure of volume to eliminate the accident year dimension".  Can I use any measure that is year specific (i.e. # of claims, # of policies)? 

[NEAS: One measure is deflated payroll; deflate by wage inflation. You can also use number of claims or policies, since deflated payroll is hard to get.]

Second, I am not completly clear on getting back to monetary values after the regression.  I understand that the values have a lognormal distribution, and to get back to the monetary value, you use the lognormal mean, and that u is the value you are trying to convert, but where do you get o?

[NEAS: Sigma is the standard error of the regression. Estimate sigma-squared as the ESS divided by the degrees of freedom.]