March 31 - April 2, 2025, in Las Vegas, Nevada. Use code MSCUST for a $150 discount! Early bird discount ends December 31.
Register NowBe one of the first to start using Fabric Databases. View on-demand sessions with database experts and the Microsoft product team to learn just how easy it is to get started. Watch now
I have a regularly changing dataset, on which I'm trying to create a measure that computes chi squared statistics. I want a measure so the underlying data can be filtered and the statistic will be recalcualted accordingly.
The raw data is surey responses for people exiting, one row per person, that hold personal information about the individual and information about motives for exiting. I made a measure that summarizes the table by a pair of characteristics, computes the expected values, then the the mean squared variation statistic, and then sums over the rows to get the chi squared statistics.
Table = ADDCOLUMNS (
ADDCOLUMNS (
SUMMARIZE (
LeavingFactorMaster,LeavingFactorMaster[Ethnicity], LeavingFactorMaster[LeavingFactor],
"people", count(LeavingFactorMaster[id])
),
"EFx1", CALCULATE(count(LeavingFactorMaster[id]),ALL(LeavingFactorMaster[Ethnicity])),
"EFx2", CALCULATE(count(LeavingFactorMaster[id]),ALL(LeavingFactorMaster[LeavingFactor])),
"EFx3", CALCULATE(count(LeavingFactorMaster[id]),ALL(LeavingFactorMaster[LeavingFactor],LeavingFactorMaster[Ethnicity])),
"EFx4", divide( CALCULATE(count(LeavingFactorMaster[id]),ALL(LeavingFactorMaster[Ethnicity]))* CALCULATE(count(LeavingFactorMaster[id]),ALL(LeavingFactorMaster[LeavingFactor])), CALCULATE(count(LeavingFactorMaster[id]),ALL(LeavingFactorMaster[LeavingFactor],LeavingFactorMaster[Ethnicity]))
)
),
"EFx5", divide(([people]-[EFx4])^2,[EFx4])
)
So far so good, EFx5 looks about right, but testing the output, there is an issue. If zero people in a ethinc group gave a particular LeavingFactor answer then there is no row in the summarized table, however this combination can have a non zero expected value. This means that the chi squared stat is lower than it should be.
I've tried to gap fill using code like this.
table2 =
var _tabl = ADDCOLUMNS(
GROUPBY(
UNION (
SUMMARIZE (
LeavingFactorMaster,LeavingFactorMaster[Ethnicity], LeavingFactorMaster[LeavingFactor],
"pp", count(LeavingFactorMaster[id])
),
ADDCOLUMNS(
CROSSJOIN(
VALUES(LeavingFactorMaster[Ethnicity]), VALUES(LeavingFactorMaster[LeavingFactor])
),
"pp", 0)
),
[Ethnicity],[LeavingFactor],
"people1",sumx(CURRENTGROUP(),[pp])
), "EFx1", 1
)
return _tabl
This does ensure that the zero rows are added, but I can't work out any way of doing calculations on "people1" to compute the chi-squared statistic like the first chunk of code.
So my question is am I on the right track? Is can this be tweaked to work? If not, is there an different approach that will compute the chi-squared stat correctly?
Any particular reason for not using the built-in function?
That function requires the input X the sum ofvthe squares of the variation from the expected values, which is the number I'm trying to compute.
If zero people in a ethnic group gave a particular LeavingFactor answer
Is that because there were no leavers in that group or because they left but didn't fill the survey?
You can consider using COALESCE to fill voids.
March 31 - April 2, 2025, in Las Vegas, Nevada. Use code MSCUST for a $150 discount!
Your insights matter. That’s why we created a quick survey to learn about your experience finding answers to technical questions.
Arun Ulag shares exciting details about the Microsoft Fabric Conference 2025, which will be held in Las Vegas, NV.
User | Count |
---|---|
23 | |
16 | |
12 | |
9 | |
7 |
User | Count |
---|---|
38 | |
32 | |
25 | |
12 | |
11 |