MovingCorr
The MovingCorr function calculates the Pearson correlation coefficient, also known as the bivariate correlation, of two numerical columns within a moving window.
Syntax
MovingCorr([Number Column 1], [Number Column 2], above, below)
Function Arguments:
- [Number Column 1] (required) - The column representing the dependent data.
- [Number Column 2] (required) - The column representing the independent data.
- above (required) - The first row to include, counting backward from the current row.
- below (optional) - The last row to include, counting forward from the current row. Defaults to 0 (current row will be the last row included).
When using this function without a sort enforced, there can be unexpected results. In order to ensure that the values are stable, verify that there is a sorted column within the table.
Example
A table of sales data lists quantity sold and total sales by day. You can use the MovingCorr function to find the relationship between the total sales of an item and the amount of an item that's purchased within several types of moving windows.
MovingCorr([Weekly Quantity Sold], [Total Weekly Sales], 4)
Here, the above argument is 4, so the correlation between [Total Weekly Sales] and [Total Quantity Sold] will be computed for the previous 4 weeks. Since the below argument was not specified, it defaults to 0 and no future weeks will be considered.
MovingCorr([Weekly Quantity Sold], [Total Weekly Sales], 0, 4)
Here, the above argument is 0, so there will not be any previous weeks included in the calculations. The below average is 4, therefore the moving correlation will be computed for each week along with the next 4 weeks.
MovingCorr([Weekly Quantity Sold], [Total Weekly Sales], 2, 2)
Here, the above argument is 2, so the previous two weeks will be included in the calculation. In addition, the below argument is 2, so the following two weeks will be included as well.
MovingCorr([Weekly Quantity Sold], [Total Weekly Sales], 8, -4)
Here is an example where the below parameter is negative. The below parameter can be negative as long as the value is less than that of the above parameter. In this example, each window begins 8 weeks before the current week and ends 4 weeks before the current week, inclusive.
This function is not currently supported on Snowflake or Redshift connections.