Tableau Correlation and Regression

Trend Lines: A trend line can provide a statistical definition of the relationship between two numerical values

Graph: Profit over Sales. Where profit is a dependent measure based on Sales measure. Each circle represents a Store.

Rt. Click on Trend Lines > Describe Trend Model

Smaller the p-value, the more significant the model and higher the correlation.

Customized Trend Lines:  Rt. Click trend Line > Edit Trend Lines

Correlation Coefficient:

The correlation coefficient is defined as:

Where ‘n’ measurements of X and Y are written as Xi and Yi (i = 1, 2 .... n) and X Tableau equivalent formula (Link:http://community.tableausoftware.com/thread/111548)

WINDOW_SUM((SUM([Sales])-WINDOW_AVG(SUM([Sales]))) * (SUM([Profit])-WINDOW_AVG(SUM([Profit])))) / SQRT(WINDOW_SUM(SQUARE(SUM([Sales])-WINDOW_AVG(SUM([Sales]))))*WINDOW_SUM(SQUARE(SUM([Profit])-WINDOW_AVG(SUM([Profit])))))

Multi-Regression: Graph below shows the linear regression of Profit over Sales for each of the Regions. Based on the Coefficient of Correlation we can analyse the following:

• East is highly positively correlated (profit grows as sales grows)
• West is highly negatively correlated (profit declines as sales grow)
• Central, South and North are in order of correlation. With least correlation for North

References/Detailed Articles: 

Wiki: http://en.wikipedia.org/wiki/Correlation_and_dependence
Tableau: http://kb.tableausoftware.com/articles/knowledgebase/statistics-finding-correlation
Trend: http://onlinehelp.tableausoftware.com/current/pro/online/windows/en-us/trendlines_significance.html