Calculating fractiles in Qlik

Every Friday at Bitmetric we’re posting a new Qlik certification practice question to our LinkedIn company page. Last Friday we asked the following Qlik Data Architect certification practice question about calculating fractiles in Qlik.

It seems like a lot of you have spend a great time around Christmas, since we are normally used to more answers.

The correct answer is B

Fractile

Fractile is a function which allows you to calculate a fractile of a set of values in a field. A fractile is a value that divides a set of data into intervals. For example, the 50th fractile of a set of values would be the value that divides the set into two equal parts: the values below the 50th fractile and the values above it. The syntax for Fractile is as follows:

Fractile([{SetExpression}] [DISTINCT] [TOTAL [<fld{, fld}>]] expr, fraction)

To illustrate how fractiles work please see the diagram below:

Here we can see that the 50th fractile is the exact median of the range. The first quartile will be the first 25% of values and so on for each quartile and the interesting to focus on is the deciles. This is all the values divided in parts of ten. So in order to get to that top 10% we need to use 0.9 as the fraction in the syntax. This will give us the value at the 90th percentile. Everything above this will be the top 10%.

AGGR

This brings us to the second part of the answer. To properly calculate the value of the 90th percentile, we need to use another aggregation. Fractile needs to be calculated on a dimension. In this case per person. So by using the Aggr function we first calculate the following (if you need to refreshen the Aggr function, please refer to this blog post):

Aggr(Sum(Nice) - Sum(Naughty), Person)

This calculates per person the score of each individual. Now it is possible to wrap this in the fractile expression to give the end result of the 90th fractile for all values of the score per person.

Use cases

Decile Analysis

One very interesting feature is to do decile analysis. Decile analysis is a useful tool for understanding how a particular variable is distributed within a dataset and for identifying trends or patterns in the data. It is done by dividing the dataset in ten equal parts (deciles) and then calculate various statistics on these. It is often used in finance and economics to analyze income or wealth distribution. Below is an example of a decile analysis for sales per car brand:

This gives great insights in revenue and which brands are responsible for the majority of revenue. This chart is made by using the following expression as a calculated dimension, so it it possible to also select the deciles.

=Aggr(
IF(Sum(Revenue) <= Fractile(total Aggr(Sum(Revenue), brandName), 0.1), 10,
IF(Sum(Revenue) <= Fractile(total Aggr(Sum(Revenue), brandName), 0.2), 9,
IF(Sum(Revenue) <= Fractile(total Aggr(Sum(Revenue), brandName), 0.3), 8,
IF(Sum(Revenue) <= Fractile(total Aggr(Sum(Revenue), brandName), 0.4), 7,
IF(Sum(Revenue) <= Fractile(total Aggr(Sum(Revenue), brandName), 0.5), 6,
IF(Sum(Revenue) <= Fractile(total Aggr(Sum(Revenue), brandName), 0.6), 5,
IF(Sum(Revenue) <= Fractile(total Aggr(Sum(Revenue), brandName), 0.7), 4,
IF(Sum(Revenue) <= Fractile(total Aggr(Sum(Revenue), brandName), 0.8), 3,
IF(Sum(Revenue) <= Fractile(total Aggr(Sum(Revenue), brandName), 0.9), 2, 1))))))))), brandName)

Other things to notice

• By default the fractile will be calculated over the used selections, you can use set expressions to change this.

• Use DISTINCT to calculate the fractiles over distinct values

• Use TOTAL to calculate fractiles over all possible values regarding the current selections.

• Nested calculations are not possible with use of Fractile() unless you use the Aggr() function.

• Fractile() is a chart function. In the script you can use FractileExc in combination with a Group By clause.

That’s it for this week!

More from the Bitmetric team