Automatic Prediction Limit Analysis
Automatic Inter-Well Prediction Limit Analysis (Script Command)
See Also:
Automatic Statistical Analysis
Processing for the Automatic Prediction Limit Analysis script command follows this sequence:
- Transform data to the original concentrations, and replace non-detects with 1/2 of the detection limit.
- If there are greater then 90% non-detects, the Poisson prediction limit test is applied. If there are greater than 15% non-detects, the non-parametric prediction limit is applied. Otherwise, normality is tested.
- If there are greater than 50 measurements, the Shapiro-Francia test for normality is applied. If there are fewer than 50 measurements, the Shapiro-Wilks test for normality is applied. If the normality passes, a parametric prediction limit test is applied. If the normality test fails, the data set is log-transformed, and step 3 is applied to the natural logarithm of the data.
- If a parametric test is appropriate, the parametric prediction limit test is applied to the original data. Otherwise the non-parametric prediction limit or Poisson prediction limit is applied according to the percentage of non-detects.