I am very new to Business Objects, having just finished the 4-day universe design course yesterday. We will be designing for WEB-I, and will need to design some reports that compare measures across two different groups and flag those that have p-values <0.05. For example, we will create reports that compare various measures for Hospital A against various measures for a group of hospitals we’ve determined are similar to Hospital A (e.g., average length of hospital stay by month, quarter, or year). We want to show hospital A’s measures side-by-side against its comparison group measures, flagging statistically significant differences in particular time periods so that hospital A can identify areas that potentially need improvement.
From what I learned in the Universe Design course, it’s clear this capability is not present in the standard product. It seems we have two options: (1) do the statistical tests completely outside BO, store the aggregate data and associated p-values in the database; (2) use the SDK to write a program using JSP or ASP in which we will pass the SQL record set out, perform the statistical test, and pass the p-values back in.
Am I on the right track with this thinking? If so, (2) would be our preferred option. With (1), the user would only receive static reports and wouldn’t be able to continue to drill down across dimensions by variables not included in the report, with p-values recalculated at each level (e.g., to see if certain groups of patients–medical, surgical, pediatric–have longer lengths of stay). With (2), I am thinking we would be able to give them p-values at each drill operation, if we programmed it so that we called the program to perform the statistical test each time the user performed a drill operation (taking into account the dimensions selected in the drill operation).
I don’t know JSP or ASP–is it even feasible to use the SDK for something like this?
Any advice, comments, observations, would be greatly appreciated!
Tracy134 (BOB member since 2004-07-23)