BOXIR2 Infoview, Oracle 10g - Percentiles

[hsandwick]

Hello:

We are building 25th/50th, 75th and 90th percentile functions within a webi report. After some initial research on the internet, I have determined that BO appears to default to the "Empirical Distribution Function - Interpolation" Percentile Type. This determines percentiles based on a range of numbers. I am curious to know what range BO uses as its default for percentiles. Is it the whole range of numbers, or a sub-range, for example?

We have inherited an historical application, the requirements for which include exactly interpreting the logic that was put in place many years ago. It appears that the historical percentile functionality is based on the "Empirical Distribution Function - Averaging" Percentile Type. Is there a way to interpret our results to match the historical results, i.e., a quick fix in BO? Otherwise, it will mean breaking down the structure of the numerical results and creating some quite complex formulae, which we are trying to avoid.

Thank you in advance for your help.

Helen

 

[cmmrfrds]

Oracle has a robust set of Analytical Functions including calculating percentiles. I would do this in Oracle not Business Objects.

Here is a link to some Oracle Functions.

http://download-west.oracle.com/docs/cd/B19306_01/server.102/b14223/analysis.htm#i1007261

 

[hsandwick]

Thank you, cmmrfrds, for your quick response, and I appreciate your pointing me to this add'l resource. When I look at Oracle's "Inverse Percentile" example, however, it, too, seems to utilize Linear Interpolation, which is also the default Percentile type in Infoview.

PERCENTILE_CONT is described as linear interpolation, using the average of 2 values if the number is even (is this really the same as Empirical Distribution Function - Averaging, just another naming convention for the same thing?), and PERCENTILE_DISC is not defined as linear interpolation, but just takes the lower of the 2 values if the number is even. So, I'm unclear as to what PERCENTILE_DISC's functionality is based upon. I guess testing in Oracle will be the only way to clarify this.

My testing of BO's percentile functionality finds that it's holding up well. However, I just want to be able to interpret my results form Empirical Distribution Function -Linear Interpolation to Empirical Distribution Function - Averaging percentile types.

We're trying to avoid the necessity to create multiple Oracle queries to capture many unique percentile scenarios among our ever-expanding list of reports - simply to address this situation - when perhaps there's an easier way.

I think there must be a way to do this, and will keep searching. If I find out any more, I will definitely post an update.

Cheers,
Helen