topic Re: Correlation between multiple columns in multiple tables (3 types of calculation involved) in DAX Commands and Tips

Correlation between multiple columns in multiple tables (3 types of calculation involved)

Peru123 — Tue, 17 Jan 2023 06:44:02 GMT

Hi all,

I need help with correlation table creation.

[Goal]

To create a correlation table like below to see the correlation (*3) between Active Return (*2) of Topix-17 Series and Monthly Return (*1) of several Indexes (trying with S&P500 to begin with).

As the data to be added regularly, I wish to do the calculation on Power BI instead of Excel.

Below is the output I made on Power BI with Excel calculated data.

X-Axis: Topix-17 Series

Y-Axis: Several Indexes (incl.S&P500)

[Indexes (Data)]

Columns and the number of rows are the same for every index (table), Total = 19 Indexes (tables)

Topix-17 Series: 17 Indexes by industry categories. 17 historical month-end price tables.

Topix Total Return Index
S&P500 Total Return Index

*1 Monthly Return = Monthly percentage change in ‘Price’

:Monthly return to be calculated for all the indexes (19 indexes)

*2 Active Return = ‘Monthly Return of Each Topix-17 Index’ – ‘Monthly Return of Topix Total Return Index’

*3 Correlation: between ‘Active Return’ of each index of Topix-17 Series & ‘Monthly Return’ of S&P500 Total Return Index

[Excel Calculation]

Hope this would explain the logics of calculations I would like to do on Power BI

Monthly Return:

Active Return:

Correlation:

[Methods Tried & Issues]

1. Inserting Index column to shift the Price value by one row

Issue: This method couldn’t put all 17 correlation values into one column, so the visual output cannot be how I want.

Solution required: If the correlation values could be calculated in one column, the output can be what I want.

2. Append the 17 tables (Topix-17 series) into a new table

(Creating a PreValue column by shifting the Prive value by one row)

Issue: Cannot shift the Price value by one row as the bottom row for each Index would calculate wrong

Solution required: if the Monthly return could be calculated without creating the PreValue column, this method may work

If anyone of you have ideas on the above, please tell me your ideas.

Thank you all in advance🙏

Peru

Re: Correlation between multiple columns in multiple tables (3 types of calculation involved)

Peru123 — Fri, 27 Jan 2023 08:57:56 GMT

here is the updates on the progress

[Update of the Progress]

I am using the 2nd method I have mentioned (2. Append the 17 tables (Topix-17 series) into a new table), and the issue I have mentioned is solved.

Syntax for each colum is below:

Column 'PreValue'

PreValue = CALCULATE(MAX('TOPIX-17 Series'[Price]), ALL('TOPIX-17 Series'), 'TOPIX-17 Series'[Index Name] = EARLIER('TOPIX-17 Series'[Index Name]), 'TOPIX-17 Series'[Index] = EARLIER('TOPIX-17 Series'[Index])+1 )

Column 'Monthly Return'

Monthly Return = IF('TOPIX-17 Series'[PreValue] = BLANK(), BLANK(), ('TOPIX-17 Series'[Price]/'TOPIX-17 Series'[PreValue])-1)

Column 'Active Return'

Active Return = [Monthly Return] - RELATED('Topix_TRN'[Monthly Return])

How the table looks like:

Column 'Correlation_S&P'

Correlation_S&P = VAR __CORRELATION_TABLE = VALUES('TOPIX-17 Series'[Index Name]) VAR __COUNT = COUNTX( KEEPFILTERS(__CORRELATION_TABLE), CALCULATE( SUM('S&P500_TRN'[Monthly Return]) * SUM('TOPIX-17 Series'[Active Return]) ) ) VAR __SUM_X = SUMX( KEEPFILTERS(__CORRELATION_TABLE), CALCULATE(SUM('S&P500_TRN'[Monthly Return])) ) VAR __SUM_Y = SUMX( KEEPFILTERS(__CORRELATION_TABLE), CALCULATE(SUM('TOPIX-17 Series'[Active Return])) ) VAR __SUM_XY = SUMX( KEEPFILTERS(__CORRELATION_TABLE), CALCULATE( SUM('S&P500_TRN'[Monthly Return]) * SUM('TOPIX-17 Series'[Active Return]) * 1. ) ) VAR __SUM_X2 = SUMX( KEEPFILTERS(__CORRELATION_TABLE), CALCULATE(SUM('S&P500_TRN'[Monthly Return]) ^ 2) ) VAR __SUM_Y2 = SUMX( KEEPFILTERS(__CORRELATION_TABLE), CALCULATE(SUM('TOPIX-17 Series'[Active Return]) ^ 2) ) RETURN DIVIDE( __COUNT * __SUM_XY - __SUM_X * __SUM_Y * 1., SQRT( (__COUNT * __SUM_X2 - __SUM_X ^ 2) * (__COUNT * __SUM_Y2 - __SUM_Y ^ 2) ) )

The syntax does not have errors, but the column does not show any value or error.

If you have ideas how to solve this, please help me🙏

Thank you all in advance

Peru

Re: Correlation between multiple columns in multiple tables (3 types of calculation involved)

Peru123 — Mon, 30 Jan 2023 06:11:09 GMT

FYI, below is how the tables are interrelated