How to calculate Kurtosis for each date

[QlikArno]

Hi Qlik community,

I want to check if my data set is normal distributed using Qlik Sense. For this I searched for some common tests like Kolmogorov–Smirnov test or Shapiro–Wilk test. I don’t think they are available in Qlik Sense!? Please inform me if these or similar tests are available. That would be very helpful.

As an alternative I want to calculate the Kurtosis and Skew. I don’t manage to calculate the Kurtosis and Skew for each row individually. Using the code

Kurtosis(total {$<DateId={">=$(vValidFrom)<=DateId"}>} aggr(sum(Value),DateId))

I get this table:

DateId	Sum(Value)	Kurtosis
		-0.25385
20191201	100	-0.25385
20191202	90	-0.25385
20191203	105	-0.25385
20191204	110	-0.25385
20191205	95	-0.25385
20191206	100	-0.25385
20191207	100	-0.25385
20191208	95	-0.25385
20191209	85	-0.25385
20191210	105	-0.25385
20191211	110	-0.25385
20191212	105	-0.25385

Here the variable vValidFrom is the date after which I want to use the data set. The data set itself consists of values summed for each date. I tried several modification of this code, removing total for example.

The Kurtosis in the example above is calculated over the entire data set. I however want to have the Kurtosis calculated for each row using the Sum(Values)’s that exist upto that point in time. For example, on the date 20191207 the Kurtosis should be based on the Sum(Value)’s from 20191201 until 20191207. This table should look like:

DateId	Sum(Value)	Kurtosis
		-0,25385
20191201	100
20191202	90
20191203	105
20191204	110	0,342857
20191205	95	-1,2
20191206	100	-0,3
20191207	100	0,312
20191208	95	0,146492
20191209	85	0,016827
20191210	105	-0,1518
20191211	110	-0,40439
20191212	105	-0,25385

How can I obtain the second table in QlikSense? Thank you in advance for your help!

[Anil_Babu_Samineni]

I don't think this will work, But as per principle of algorithm you can try this way (JFH)

((Count(Value)*(Count(Value)+1)/((Count(Value)-1)*(Count(Value)-2)*(Count(Value)-3)))*
Sum(Aggr(((Min(Aggr(Avg(Value), DateId))-avg(Value))*4)/(Stdev(Value)*4), DateId)))-
(((Count(Value)-1)*2)/(Count(Value)-2)*(Count(Value)-3))

[QlikArno]

Thank you for your reply. Your suggestion did not solve the issue. In an example I get values of < -70. These are very low since the kurtosis of a normal distribution is between -3 and 3. Thanks for the effort though!

Other suggestions on how to test if a data set is normal distributed are very welcome.

[lorenzoconforti · Accepted solution]

You need to use the RangeKurtosis function together with the Above function

RangeKurtosis(Above( sum(Value), 0, RowNo()))

See attached

[QlikArno]

Thank you for your reply! I didn't try RangeKurtosis in combination with Above before. It solved my issue.