Calculating column values based on date match between tables

[Anonymous]

I'm attempting to perform a column calculation that will return a value based on the last date of an instrument calibration prior to a quality control check. I have two tables, Calibration and QC, which are related by instrument serial number. For each instrument, there may be multiple calibration dates and quality control checks, during which a result (Xq) is generated. The calculated column (cal check) should compare the difference between the QC check Xq and the Xq result of the last calibration prior to that check. 

Example:

Calibration

Serial	Cal Date	Cal Xq
100141	12/5/19 14:12	0.081
100141	4/12/15 11:36	0.085
100290	6/16/15 9:19	0.077
100321	1/21/19 14:44	0.082
100321	1/21/19 11:12	0.079
100670	1/2/20 11:24	0.079
100670	9/4/18 15:39	0.081

QC

Serial	QC Date	Cal Date	QC Xq	Cal Check
100141	6/15/19	4/12/2015	0.081	-4.71%
100141	10/15/19	4/12/2015	0.082	-3.53%
100141	2/15/20	12/5/2019	0.079	-2.47%
100141	6/15/20	12/5/2019	0.08	-1.23%
100290	6/15/19	6/16/2015	0.079	2.60%
100290	10/15/19	6/16/2015	0.078	1.30%
100290	2/15/20	6/16/2015	0.077	0.00%
100290	6/15/20	6/16/2015	0.077	0.00%
100321	6/15/19	1/21/2019	0.08	-2.44%
100321	10/15/19	1/21/2019	0.081	-1.22%
100321	2/15/20	1/21/2019	0.079	-3.66%
100321	6/15/20	1/21/2019	0.082	0.00%
100670	6/15/19	9/4/2018	0.079	-2.47%
100670	10/15/19	9/4/2018	0.077	-4.94%
100670	2/15/20	1/2/2020	0.082	3.80%
100670	6/15/20	1/2/2020	0.081	2.53%

Fortunately, there is a helper column in the QC table that lists when the last cal was performed. Unfortunately, it is just a date with no time stamp, whereas the Cal Date column in the Calibration table does include the time, and there may be multiple entries from the same day. If I were doing this in Excel, I would use an array to match serial number and cal date between the two tables to return the appropriate Cal Xq, and calculate the difference from there. If they are sorted newest-to-oldest in the table, it will give me the correct result. But, I'm new to Power BI and haven't found a comparable method of doing this with DAX. 

Any help would be appreciated, thank you!

[jdbuchanan71]

@Anonymous 

Try this as a start.  Add this calculated column to your QC table.

Cal Check Column = 
VAR _Serial = QC[Serial]
VAR _Date = QC[QC Date]
VAR _LastCal = CALCULATE(
    MAX (Calibration[Cal Date]),
    Calibration[Serial] = _Serial,
    Calibration[Cal Date] < _Date
)
RETURN
CALCULATE(
    MAX(Calibration[Cal Xq]),
    Calibration[Serial] = _Serial,
    Calibration[Cal Date] = _LastCal
)-QC[QC Xq]

[Anonymous]

@jdbuchanan71 

That's definitely a promining start, thanks! I'm getting the error "A single value for column 'Serial' in table 'QC' cannot be determined." due to there being multiple entries in the table for each serial number. Using a MIN or MAX function yields incorrect results, as I think it's picking values associated with different dates in that case. Is there a way to specify that the variable only uses the serial number, date, and Xq from that row?

I appreciate the help!

[jdbuchanan71]

@Anonymous 

You are adding it as a calculated column on the QC table yes?  The VAR will read the current row of the table it is working on.

[richbenmintz · Accepted solution]

Hi @Anonymous ,

here are two calc Columns that I think do the trick

last Cal Xq val = 
var serial = [Serial]
var calDate = [QC Date]
var maxDate = CALCULATE(MAX('Calibration'[Cal Date]), FILTER(Calibration, Calibration[Serial] = 'QC'[Serial] && Calibration[Cal Date] <= calDate ))
return 
CALCULATE(MAX('Calibration'[Cal Xq]), FILTER('Calibration', Calibration[Serial] = serial && Calibration[Cal Date] = maxDate))

var = DIVIDE([QC Xq], [last Cal Xq val])-1 

Hope this Helps

Did I answer your question? Mark my post as a solution!
Did my answers help arrive at a solution? Give it a kudos by clicking the Thumbs Up!

[Anonymous]

Hi @richbenmintz,

Apologies for the slow reply! I had been pulled into other projects and this one got backburnered. 

I'm using your two-column solution which looks promising, but getting a circular dependency error for the second column when I attempt the DIVIDE function. 

Using the information found at https://community.powerbi.com/t5/Desktop/Row-Identifier-within-Power-BI-for-Desktop-Circular-depende... I've been trying to use ALLEXCEPT in the CALCULATE function to fix the error, but haven't had much luck.

Thanks again for your help!

[richbenmintz]

Hi @Anonymous,

Can you share your Calculated Column formula that is generating the circular dependency?

[Anonymous]

The formulas I'm using are essentially what you had suggested above; so column 1 returns the last calibration Xq as intended:

Last Cal Xq = 
VAR _SERIAL = [SerialNumber]
VAR _RPCDATE = [TestDate]
VAR _MAXDATE = 
    CALCULATE(
        MAX(RECORDS_AD[Cal Date]), 
        FILTER(RECORDS_AD, RECORDS_AD[SerialNumber] = _SERIAL && RECORDS_AD[Cal Date] <= _RPCDATE)
        )
RETURN
    CALCULATE(
        MAX(RECORDS_AD[Xq]), 
        FILTER(RECORDS_AD, RECORDS_AD[SerialNumber] = _SERIAL && RECORDS_AD[Cal Date] = _MAXDATE)
        )

But the second column I'm calling "Off" gives the circular dependency error:

Off = 
DIVIDE(RECORDS_RP[Xq], RECORDS_RP[Last Cal Xq])-1

[richbenmintz]

Hi @Anonymous,

What is the difference between the RECORDS_AD and RECORDS_RP Tables? Is RECORDS_RP[Xq] a calculation that references RECORDS_AD[Xq]?

[Anonymous]

Apologies, I should have made the language consistent with the example above - RECORDS_AD is essentially the Calibration table ("AD" for calibration "adjustment"), and RECORDS_RP is the QC table (RP for "routine performance" check); so my intent is to calculate the difference between each RECORDS_RP[Xq] and the corresponding RECORDS_AD[Xq] from the last calibration adjustment date. 

Sorry for the confusion!

[richbenmintz]

Hi @Anonymous,

Can you provide a sample pbix with the error?

[richbenmintz]

can you share your pbix