Forum Discussion
creating a dynamic table in Power BI
If the error message is about SUMMARIZECOLUMNS and ADDMISSINGITEMS not being allowed in the current context, that's an issue with SUMMARIZECOLUMNS not playing well with slicers on multiple columns.
I re-wrote the measure to use the old ADDCOLUMNS style pattern that was used before SUMMARIZECOLUMNS was introduced:
MK-Stat = VAR ComplaintsByFiscalMo = ADDCOLUMNS(VALUES(PMS_COMPLAINT[FISCAL_MON_START_DT]),"CountComplaints", CALCULATE(COUNTROWS(PMS_COMPLAINT))) RETURN SUMX(ComplaintsByFiscalMo, SUMX(ComplaintsByFiscalMo, IF([FISCAL_MON_START_DT]>EARLIER([FISCAL_MON_START_DT])&&[CountComplaints]>EARLIER([CountComplaints]), 1, IF([FISCAL_MON_START_DT]>EARLIER([FISCAL_MON_START_DT])&&[CountComplaints]<EARLIER([CountComplaints]), -1)) ) )I've also re-uploaded the change in the .pbix file on Google Drive. The same link will download the newer version.
- Hi thanks for the excellent feedback. I'll have to post another question as I have just applied your calculation in the SE for and the Z test as per link and I am getting a different p value to the one showing in the main fact table in column mannkandelY. See what se and z test values u get as per link.
OK. I'm not a statistics person, so let's break this down. If I've gotten any of the top-level bullet points wrong, please let me know.
- You want a measure to calculate the MK-stat for some data. The previous answer I gave gets the correct answer for the small example data, so now we just need to adjust it for your large dataset - Yes my data set PMS_Complaimts
- This data to be used is a subset of your data, using only data from the previous 12 months - and 24 and 36 months , this is shown by the Y and X and W letters in PMS_Financials Table
- What is the field for determining whether a row is in the last 12 months? In the PMS_COMPLAINT table, the options I see are: 'COMPLAINT_REPORTED_DT', 'REPORTED_FISCAL_YEAR_PD_SK', 'COMPLAINT_CLOSURE_DT', 'FISCAL_MON_START_DT', or is it some other field?
- The MK test seems to assume your data is in chronological order. That's fine, and probably means we want to order by the date instead of by MATERIAL_ID. - That is correct or we could use the date field as above
- Is this the same date we use to determine if a row counts as in the last 12 months, or a different one? - Yes
- It seems likely that you'll still want to group the data by MATERIAL_ID. This is fine, we just have to pick a way to aggregate the values within a given group. - Yes, and also but would it be possible to work this out based on what every filter i passed into the data set. (Seperate question)
- In the link, the values used in the MK test were just provided. We need to determine a way to calculate these values for your data so we can then use it as input for the MK test. if you scroll down i the link the calculation are there. see link again http://www.real-statistics.com/time-series-analysis/time-series-miscellaneous/mann-kendall-test/
- I see a staggering number of possibilities for what this input value could be. It could be a count of complaints on each date, it could be a sum of QTY_AFFECTED on each date, it could be an average of ComplaintsY/X/W for each date, or something completely different. - Just the count of complaints by date
- I have absolutely no idea which fields/groupings/aggregations you want to use to generate these values. - material_id, click on filter Manufacturing site, Region, Country, REPORTED_FAILURE_MD_CD and Reportability (dounut chart), just thses fields, if it can be done, allows me to work out the P-value for any combination.
- This is where I need the most input from you. If it helps, the link you shared had a line graph plotting the given values. What fields/measures would you use as the X and Y axis for a similar graph with your data? - look at the graph called Total compliants with trends.
- Once you are able to correctly calculate the MK-stat (-44 in the example data), how do you want to display it in the report? In a table visual with MATERIAL_ID? In some sort of graph? - See tooltip 1(second tab). The graph "Total complaints with trend" gives us the data point that you could put into R and check that your p-value is correct. Those are the data points for 12 months for a given product.
There we go. Now that I know what values you want input into the MK test, we can use it in the context of your report:
MK-Stat =
VAR ComplaintsByFiscalMo = SUMMARIZECOLUMNS(PMS_COMPLAINT[FISCAL_MON_START_DT],"MonthGroup", SELECTEDVALUE(PMS_COMPLAINT[FISCAL_MON_START_DT]),"CountComplaints", COUNTROWS(PMS_COMPLAINT))
RETURN
SUMX(ComplaintsByFiscalMo,
SUMX(ComplaintsByFiscalMo,
IF([MonthGroup]>EARLIER([MonthGroup])&&[CountComplaints]>EARLIER([CountComplaints]), 1,
IF([MonthGroup]>EARLIER([MonthGroup])&&[CountComplaints]<EARLIER([CountComplaints]), -1))
)
)If the error message is about SUMMARIZECOLUMNS and ADDMISSINGITEMS not being allowed in the current context, that's an issue with SUMMARIZECOLUMNS not playing well with slicers on multiple columns.
I re-wrote the measure to use the old ADDCOLUMNS style pattern that was used before SUMMARIZECOLUMNS was introduced:
MK-Stat = VAR ComplaintsByFiscalMo = ADDCOLUMNS(VALUES(PMS_COMPLAINT[FISCAL_MON_START_DT]),"CountComplaints", CALCULATE(COUNTROWS(PMS_COMPLAINT))) RETURN SUMX(ComplaintsByFiscalMo, SUMX(ComplaintsByFiscalMo, IF([FISCAL_MON_START_DT]>EARLIER([FISCAL_MON_START_DT])&&[CountComplaints]>EARLIER([CountComplaints]), 1, IF([FISCAL_MON_START_DT]>EARLIER([FISCAL_MON_START_DT])&&[CountComplaints]<EARLIER([CountComplaints]), -1)) ) )I've also re-uploaded the change in the .pbix file on Google Drive. The same link will download the newer version.
HI
do i add this as new table or a measure into PMS_Complaints table????
sorry to be a pain
but i cannot get this to work, can you kindly upload the pbix, please.
I created this as a measure in your _Measures table.
Here's the .pbix file. The measure is MK-Stat, and I've displayed it in a card in the bottom right, just under MATERIAL_ID_FREQUENCY
- Hi I did the same and it produced and error. Hmmmm.
What error are you receiving?
- Hi thanks for the excellent feedback. I'll have to post another question as I have just applied your calculation in the SE for and the Z test as per link and I am getting a different p value to the one showing in the main fact table in column mannkandelY. See what se and z test values u get as per link.
- Hi silly question pms complainants Fiscal month st date is this for 12 month or 24 or 36.....what period.
I'm able to get the same p-values, though I had to set up a new measure that mimicked the calculation the website was doing to get the value in P19. They only match when you don't have any filters applied. I assume this is how they were originally calculated.
Also, due to an overflow error, the p-value can't be displayed in the Product Complaints Totals with Trend Direction matrix, so I displayed it in a final card. You can check that it's getting the correct value by selecting a single product from the matrix.
I've re-uploaded the new file to GDrive, feel free to test it out!
You tell me. I figured it was just the start date of the current fiscal month that the complain was registered, and that if you changed your filter, it would include values from other periods.
So it should apply to whatever the filtered period is I guess? I haven't messed with however you set that up.
Hi Mcmahan...
did you get my post on the frequncy question?