Qlik Community

Ask a Question

New to QlikView

If you’re new to QlikView, start with this Discussion Board and get up-to-speed quickly.

Announcements
Welcome to our newly redesigned Qlik Community! Read our blog to learn about all the new updates: READ BLOG and REPORTED ISSUES
cancel
Showing results for 
Search instead for 
Did you mean: 
Contributor
Contributor

Ignore outliers in a graph

Hello everyone,

I'm wondering whether it is possible to ignore statistical outliers (fe. >2*standarddeviation (alternatives are welcome as well)) from the calculation of an performance graph.

Formula below currently calculates the blue line shown above.

AVG(aggr((Sum(Aggr(fabs(Date(avg(TOTAL <[inkooporderregel link]>
Aggr(
If(volgnrhist = Max(TOTAL <[inkooporderregel link]> volgnrhist),Datum_uiterlijke_leverdatum),
[inkooporderregel link],volgnrhist
))
  -
aggr(max(Datum_ontvangst),[inkooporderregel link])),[inkooporderregel link]))),[inkooporderregel link]))

It represents the absolute deviation in actual delivery date in comparisson to the last confirmed delivery date

   

LineConfirmed delivery dateActual delivery dateAbsolute deviation
115-02-1819-02-184
215-02-1819-02-184
315-02-1819-02-184
415-02-1819-02-184
523-02-1819-02-184
613-03-1819-02-1822

The Average deviation would be 7 and standarddeviation is 7.34. In case where I would see deviations exceeding the average by 2*standarddeviation as outliers I would see 22 as an outlier and the average with the outlier not taken into acoount is 4.

Thus, Can anyone help me to ingore outliers in such a calculation?

Note: I can not work with a min level or max level (it should be a statistical outlier)

Many thanks in advance

Thomas

3 Replies

It might be easy to test out if you can share a qvw?

Contributor
Contributor

Hereby a sample of the data needed.

Since I don't understand the data in your application, I am having a hard time knowing what the expected output needs to be for your sample... Would you be able to explain what rows needs to be excluded from the line chart based on Avg +/- 2 Stdev?