Updated: 31 Oct 2012

Use MovingSTDEV to calculate the sample standard deviation of column values in an ordered resultant table, without the need for a self-join. The sample standard deviation is calculated for each value from the first value in the window to the last value in the window. If the column values are presented to the functions out of order, an error message will be generated.

SELECT [Example].[wct].[MovingSTDEV](

<@Val, float,>

,<@Offset, int,>

,<@RowNum, int,>

,<@Id, tinyint,>)

GO

the value passed into the function. *@Val* is an expression of type **float** or of a type that can be implicitly converted to **float**.

specifies the window size. *@Offset* is an expression of type **int** or of a type that can be implicitly converted to **int**.

the number of the row within the group for which the sum is being calculated. If *@RowNum* for the current row in a set is less than or equal to the previous *@RowNum* and *@RowNum* is not equal to 1, an error message will be generated.* @RowNum* is an expression of type **int** or of a type that can be implicitly converted to **int**.

a unique identifier for the MovingSTDEV calculation. *@Id* allows you to specify multiple MovingSTDEV calculations within a resultant table. *@Id* is an expression of type **tinyint** or of a type that can be implicitly converted to **tinyint**.

· If *@Id* is NULL then *@Id* = 0.

· To calculate the running standard deviation from the beginning of a dataset or a partition, use the RunningSTDEV function.

· If *@RowNum* is equal to 1, MovingSTDEV is equal to NULL

· *@RowNum* must be in ascending order.

· There may be cases where the order in which the data are returned to the function and the order in which the results are returned are different, generally due to parallelism. You can use OPTION(MAXDOP 1) or OPTION(MAXDOP 1,FORCE ORDER) to help eliminate this problem.

In this example, we have 20 rows of sample data from a population with a mean of 100 and a standard deviation of 15 and we want to calculate the sample standard deviation for the previous 4 row and the current row in a window where the x-values are sorted in ascending order.

SELECT cast(x as money) as x

,cast(wct.MovingSTDEV(x,4,ROW_NUMBER() OVER (ORDER BY x ASC),NULL) as money) as [STDEV]

FROM (

SELECT 1,85.2968 UNION ALL

SELECT 2,88.2566 UNION ALL

SELECT 3,100.1934 UNION ALL

SELECT 4,116.3052 UNION ALL

SELECT 5,109.6867 UNION ALL

SELECT 6,130.3847 UNION ALL

SELECT 7,76.5458 UNION ALL

SELECT 8,99.5511 UNION ALL

SELECT 9,101.5546 UNION ALL

SELECT 10,114.318 UNION ALL

SELECT 11,100.2686 UNION ALL

SELECT 12,110.5982 UNION ALL

SELECT 13,91.4181 UNION ALL

SELECT 14,118.5804 UNION ALL

SELECT 15,126.6649 UNION ALL

SELECT 16,103.8977 UNION ALL

SELECT 17,82.2819 UNION ALL

SELECT 18,123.3369 UNION ALL

SELECT 19,98.9415 UNION ALL

SELECT 20,89.1731

) s(rn,x)

This produces the following result.

x STDEV

--------------------- ---------------------

76.5458 NULL

82.2819 4.056

85.2968 4.4455

88.2566 5.0014

89.1731 5.1136

91.4181 3.5558

98.9415 5.1454

99.5511 5.403

100.1934 5.1561

100.2686 3.7598

101.5546 0.9742

103.8977 1.7281

109.6867 3.9646

110.5982 4.7057

114.318 5.1934

116.3052 4.7821

118.5804 3.759

123.3369 4.7597

126.6649 5.0838

130.3847 5.7536