SQRT error on PowerBI/PowerPivot when using column references

[will1329]

Hi,

I am attempting to do a workaround for the lack of NORMSINV function in PowerBI/PowerPivot. Part of my formula is:

B = 1.96*SQRT(1.96^2+4*[NUMERATOR]*(1-[NUMERATOR]/[DENOMINATOR]))

When NUMERATOR = 1 and DENOMINATOR = 1, I should get a result of 3.84.

However, in PowerBI/PowerPivot this result in the error:

An argument of function 'SQRT' has the wrong data type or the result is too large or too small.

The error is not present when using the exact same formula in excel and also goes away when I replace the column references with values (i.e. if I replace both NUMERATOR and DENOMINATOR with 1 I get the desired result).

Does anyone know why this might be happening?

Many thanks,

Will

[Greg_Deckler · Accepted solution]

The documentation on SQRT states that it returns an error for negative numbers. You are most likely getting negative numbers passed to the SQRT function. To test, try creating this column:

D = VAR myvalue = [NUMERATOR]/[DENOMINATOR]
    VAR myvalue2 = 1.96^2
    VAR myvalue3 = myvalue2+4*[NUMERATOR]*(1-myvalue)
    RETURN myvalue3
//    RETURN 1.96*SQRT(myvalue3)

If you see negative numbers, you have your answer. See documentation here:

https://msdn.microsoft.com/en-us/library/ee634559.aspx

Note the Remarks section.

You could potentially fix this by wrapping the inside of the SQRT with an ABS. You could create a VAR to see if the value that you are taking the SQRT of is negative and, if so, multiply by -1 outside of the SQRT if you really need the negatives.

[Dave1 · Accepted solution]

Hi, 

The issue is caused by negative results for SQRT function.

Create a measure for your numerator = sum(YourNumerator)

Create a measure for your denominator =  sum(YourDenominator)

Test logic

    Measure = Measure = 1.96*SQRT(1.96^2+4*1*(1-1/1))    === 3.84

Prod Code

    Measure2 = Calculate(1.96*SQRT(ABS(1.96^2+4*[sum1]*(1-[sum1]/[Sum2]))))

Measure and Measure 2 produce the same result for Field A, where Numerator and denominator are 1.

[Greg_Deckler · Accepted solution]

The documentation on SQRT states that it returns an error for negative numbers. You are most likely getting negative numbers passed to the SQRT function. To test, try creating this column:

D = VAR myvalue = [NUMERATOR]/[DENOMINATOR]
    VAR myvalue2 = 1.96^2
    VAR myvalue3 = myvalue2+4*[NUMERATOR]*(1-myvalue)
    RETURN myvalue3
//    RETURN 1.96*SQRT(myvalue3)

If you see negative numbers, you have your answer. See documentation here:

https://msdn.microsoft.com/en-us/library/ee634559.aspx

Note the Remarks section.

You could potentially fix this by wrapping the inside of the SQRT with an ABS. You could create a VAR to see if the value that you are taking the SQRT of is negative and, if so, multiply by -1 outside of the SQRT if you really need the negatives.

[will1329]

Hi,

Thanks for the help both. Thought I'd made sure there were no negatives but a few had slipped past as the records contained blanks (12 out of 15million isn't too bad I think!). I've added your useful suggestions to deal with negatives for extra certainty. Just goes to show you should check your data before you analyse it

Thanks