Excluding Outlier Data from Calculations

[jhomrighaus]

Hello,
I am attempting to generate a report to calculate a set of statistical control limits from QC data contained in an external database.

I can aquire the data without issue and I can calculate all the of the various statistics without issues, where I am encountering difficulty is that I want to then filter out all of the data that meets the qualification for an outlier and recalculate the limits, but I can't seem to find a way to do it.

Any ideas?

 

[beltmanjr]

The information provided is probably not sufficient to provide a good answer however, why not simply suppress the details, sections where the criteria for the outlier is satisfied?

 

[jhomrighaus]

I can do that the problem is the summary formulas still use the data to calculate even though it is suppressed in the details section.

 

[beltmanjr]

Of course, that is the beauty about the program
To prevent that use running totals or custom formula calculating numbers, or the sum function with criteria

 

[lbass]

For more help, you would have to define what you mean by "outlier" data in mathematical terms.

-LB

 

[jhomrighaus]

In this case I import the date which consists primariy of %recovery of expected results. This is groupped by The test and further by the type of extraction used. Then finally it is grouped by the type of QC(duplicate, control etc) on each of these groups you calculate the Average and Standard Deviation of the all the records in the group, outliers are identified as any value which is outside of 3 standard deviations from the mean for the data set.

I am drawing from an Advantage ODBC connection to a set of tables on my local machine, I am running a complex 7 part union query to draw in data from a series of different log tables within the database. This is done using a SQL command in the Database manager window.

I would like to be able to run a report of of the results by filtering out the outliers and recalculating the limits(which may result in additional outliers of course)