Overload List
| # | Signature | Description |
|---|---|---|
| 1 | procedure StdDev(const SumOfSquares: TMtx; const Sum: TMtx; Averages: Integer; const aResult: TMtx); | Return standard deviation in aResult. |
| 2 | procedure StdDev(const SumOfSquares: TVec; const Sum: TVec; Averages: Integer; const aResult: TVec); | Return standard deviation in aResult. |
| 3 | function StdDev(SumOfSquares: Double; Sum: Double; Averages: Integer): Double; | Result = the sample standard deviation from running sums |
Overload 1: procedure StdDev(const SumOfSquares: TMtx; const Sum: TMtx; Averages: Integer; const aResult: TMtx);
Return standard deviation in aResult.
| # | Name | Type | Description |
|---|---|---|---|
| 1 | SumOfSquares | TMtx | |
| 2 | Sum | TMtx | |
| 3 | Averages | Integer | |
| 4 | aResult | TMtx |
Result: stored in self (calling object)
Overload 2: procedure StdDev(const SumOfSquares: TVec; const Sum: TVec; Averages: Integer; const aResult: TVec);
Return standard deviation in aResult.
| # | Name | Type | Description |
|---|---|---|---|
| 1 | SumOfSquares | TVec | |
| 2 | Sum | TVec | |
| 3 | Averages | Integer | |
| 4 | aResult | TVec |
Result: stored in self (calling object)
Overload 3: function StdDev(SumOfSquares: Double; Sum: Double; Averages: Integer): Double;
Compute the sample standard deviation from running sums.
| # | Name | Description |
|---|---|---|
| 1 | SumOfSquares | The accumulated sum of squares sum x_i^2. |
| 2 | Sum | The accumulated sum sum x_i. |
| 3 | Averages | The sample count n, which must satisfy n>1. |
Returns: Double - The unbiased sample standard deviation sigma.
Remarks:
Returns the unbiased (Bessel-corrected) sample standard deviation computed from the precomputed sum of squares and sum of the samples: sigma = sqrt((n sum x_i^2 - (sum x_i)^2)/(n (n-1)))
n*Sum(x^2) - Sum(x)^2
StdDev = ( ------------------------- )^0.5
n*(n-1)
n... number of averages taken
Sum(x^2)... sum of squares of samples
Sum(x)... sum of samples
where n = Averages is the number of samples, sum x_i^2 = SumOfSquares and sum x_i = Sum. The argument n must be greater than 1. When severe cancellation makes the radicand negative (catastrophic loss of precision for large n) the routine raises an exception rather than returning NaN.