Matrix sub-total calculation?

Sorry but I think you didn't get my point. To fix an issue you need to go step by step with your measures and test them one by one until you get to that issue.

For instance the measure "Upside (Graph)" is full of variables and none of them makes sense to me therfore how do you know that the values in the Matrix are accurate and if so what makes the total coming out wrong???

Please re read the remarks in the file I sent you and follow the actions i made to get to the confirmation that "MY TABLE" was close to your inteded income.

Just to add, it is good to use Variables but it's a must that you know what those variables are going to do and how to use them because sometimes Var don't work as intended when it is embeded inside CALCULATE(as an example). So adding simple and small measures and then combine them into one can be more accurate that using multiple VAR inside a single measure. 

Good luck

aj1973 

I think I've found the solution.  Will post here when done, but please stop looking at this for now.

aj1973 lbendlin 

Here is a cleaned up version of the PBIX: https://drive.google.com/file/d/12_k_tO8BdrXV73Lbho_whabGP-jOlasP/view?usp=sharing

Here is the current formula that's not generating the intended output:

Upside (averaged within a given sale year for graph output) = 

VAR OUTPUT = 
        AVERAGEX( 
            ADDCOLUMNS (
                SUMMARIZE (
                    ALLSELECTED ('Helivalues Transaction History'[Sale Year]),
                    'Helivalues Transaction History'[Sale Year],
                    "COUNT", COUNT ( 'Helivalues Transaction History'[Year of Build] )
                ),
                "__Output", [Upside (Graph)]  
            ),
            [__Output]
)


RETURN
    
    OUTPUT

Here is the current output:

The intended output is that the Matrix at the bottom of the above-screenshot is populated with the columnar average of the Matrix at the top of the above-screenshot.

The formula for the [measure] in the top Matrix is: 

Upside (Graph) = 

VAR EXPECTED_PRICE = 
    
    IF(
        HASONEVALUE('Helivalues Transaction History'[Year of Build])
        , [Expected (Graph)]
        , AVERAGEX('Helivalues Transaction History',[Expected (Graph)])
    )

VAR COEFF_UPSIDE = 
   
    IF(
        HASONEVALUE('Helivalues Transaction History'[Year of Build])
        , [______G1 - Coeff_Upside {equals} sqrt{power{{E},{F}}}]
        , AVERAGEX('Helivalues Transaction History',[______G1 - Coeff_Upside {equals} sqrt{power{{E},{F}}}])
    )

VAR UPSIDE_GRAPH =
    EXPECTED_PRICE * COEFF_UPSIDE


VAR OUTPUT_FINAL = 
        AVERAGEX( 
            ADDCOLUMNS (
                SUMMARIZE (
                    ALLSELECTED ('Helivalues Transaction History'[Year of Build]),
                    'Helivalues Transaction History'[Year of Build],
                    "SUM", COUNT ( 'Helivalues Transaction History'[Sale Year] )
                ),
                "__Output", UPSIDE_GRAPH 
            ),
            [__Output]
)
 

RETURN

    IF(
        HASONEVALUE('Helivalues Transaction History'[Year of Build]),
        OUTPUT_FINAL
        , AVERAGEX('Helivalues Transaction History',OUTPUT_FINAL)
    )
    

Hi mrothschild 

Before fixing the second Matrix, are you sure the values in the first Matrix are accurate?

same here

Hey mrothschild 

I spent some time to see why the totals are not accurate, here is your file where I added a new page and corrected a tiny view:

https://drive.google.com/file/d/1svcBN0vf1oEPSimJ60GibzMJaDvjhtNn/view?usp=sharing

A quick ex 

Looking at the measure it is like you are dividing  (X/Y) * X = Y all the time.

aj1973 lbendlin 

PBIX file is here; https://drive.google.com/file/d/12_k_tO8BdrXV73Lbho_whabGP-jOlasP/view?usp=sharing

Excel file with exported results from upper Matrix: https://docs.google.com/spreadsheets/d/12nChVgYinw5RMB-M5S4lv96hLQQzMwjT/edit?usp=sharing&ouid=115111367427670973715&rtpof=true&sd=true

____________________________________________

I've added a new tab, called "Troubleshooting for forum help - buildup. . . "

This indicates that the upper matrix in the original screenshot below is all "fruit from a poisonous tree" at a summarized level but not at an individual cell level, which I believe is what you were driving at aj1973 .

The data in each of the individual cells is accurate/desired/intended.  The summarized columnar average of these antecedent measures is wrong. 

The programming for each of these [measures] is all structured the same, so I'm hoping if I can fix this structure everything else will fall into place.  {A} uses STDEVX.S, but STDEVX.P also doesn't summarize to the desired output.  {B} & {C} use AVERAGEX, and {C} isn't summarizing at all, i.e., BLANK() in one of the sub-total rows, even though two individual cells are above it, and is completely off when being summarized elsewhere - a bunch of negative numbers being summarized to a positive number.

{D}, {E}, and {G} are all functions of {A}, {B}, and {C}, so unless the first 3 summarize correctly, I imagine will be hard for these to summarize correctly.

{F} is summarizing correctly.

{A, B, C, etc.} =

VAR TARGET_MEASURE=
            AVERAGEX(
                SUMMARIZE (
                        ALLNOBLANKROW('Helivalues Transaction History'),
                        'Helivalues Transaction History'[Sale Year]

                    )
                , [Average Annualized Price change (weighted by Model Year Units)]
            )
        

RETURN
        IF(
            HASONEVALUE('Helivalues Transaction History'[Year of Build])
            , TARGET_MEASURE , 
            AVERAGEX('Helivalues Transaction History', TARGET_MEASURE)
        )

More broadly, I apologize for my inability to articulate the issues clearly and the idiosyncracies in the coding.  I'm relatively new to DAX and don't have an intuitive understanding.  Most of what I've built is equivalent to duct tape and chicken wire that come from hitting a wall, going to the forums, and trial & error until I find something that works.

Thanks for your help!

aj1973 

Thank you for that.  I can confirm that "YOUR TABLE" in the PBI file your shared is the intended outcome.  That calculation is a version of SUMPRODUCT in Excel, where it's delivering the weighted-average price based on units in a given year.   

Imagine it was sneakers and we were selling some for $50 and some for $100.  If we sold twice as many $50 sneakers our "Total Sales" would be the same for each sneaker type, but the weighted average price would be $67, not $75.  At an individual row level, you would be always, in your example be dividing by Y and left with X, but at a summarized aggregated level, the math changes.

aj1973 

Unless the calculation results in something different for your file than for the one you shared with me, "YOUR TABLE" is producing the desired output, and "MY TABEL" is not.  

But you have peeled back to identify the core problem, which is the methods I've been reading about that are used to summarize a [measure] calculated using a simple average don't seem to be working for [measures] calculated using a weighted average approach.  

The following matrix In this file: https://drive.google.com/open?id=12_k_tO8BdrXV73Lbho_whabGP-jOlasP

is my attempt at a [measure] by [measure] approach.  The _TEST columns are based on simple averages and not weighted averages and the methodology to get to the desired row sub-totals is working.  

I need to figure out how to do the same for a weighted average [measure].

aj1973 

Premature, but making progress.

The new interim step measure, {A int1}  goes onto the matrix in individual cells as [Units] * [{A}] and is calculating correctly at an individual cell level.  The [Units] is calculating correctly at both individual and row-subtotal level. 

So now trying to summarize {A int1}, consistently associated with the individual cell data that is going into it.