Adjustment Settings
Adjustment settings include all settings, which are required for parameter estimation with least-squares method, including data snooping with statistical tests.
To select parameter settings in the software (ADJUST in the native DLL), the corresponding files must be stored in a specific folder:
C:\ProgramData\point electronic\microCAL\ParameterSettings
If different, the corresponding application data directory must be stated in the application settings file. The ParameterSettings folder can contain any number of files with specific transformation parameters, which can then be selected in microCAL. The file extension must be .xml; the file name without the extension is used as the display name microCAL.
File format is XML.
Main sections
| Element | Attributes | Explanation | Value | Examples | Default |
|---|---|---|---|---|---|
| iterMax | Maximum number of iterations | Integer | 20 | ||
| iterBreakKrit | Break criterium for iteration | Float | 1E08 | ||
| nvMax | Threshold for outlier test, normalized residuals | Float | 3.84 | ||
| m0apri | Global mean error a priori | ||||
| DoDataSnooping | Apply statistical test for outlier filtering | 1 = true, 0 = false | |||
| ds_first_iter |
Thresholds for outlier test (nvMax)
| k | alpha | 1-beta | nvMax |
|---|---|---|---|
| 1.96 | 99.7% | 70% | 3.52 |
| 1.96 | 99.7% | 80% | 3.84 |
| 1.96 | 99.7% | 90% | 4.28 |
Thresholds based on Baarda's outlier test (data snooping). Standard distribution is approximately assumed (which is fulfilled especially with high redundancies). For testing, normalized residuals must be applied.
- k: Quantil for normal distribution
- alpha: Type 1 error rate (false positive, also known as confidence or significance level)
- beta: Type 2 error rate (false negative)
- nvmax: Threshold for outlier test, normalized residuals
For normal distribution with 3-sigma rule following values are given:
- k = 1.96
- alpha = 99.7%