WS Node: Stability Analysis - Advanced Tab
Select the Advanced tab of the Stability Analysis node specifications dialog box to access these options.
Poolability of intercepts alpha
Level of p-value that will be deemed significant for inferring equal intercepts. The null hypothesis is that all batches have equal intercepts.
Poolability of slopes alpha
Level of p-value that will be deemed significant for inferring equal slopes. The null hypothesis is that all batches have equal slopes.
Use pooled MSE for unequal int/unequal slope model
This option is only applicable when an unequal intercept and unequal slope model is selected as the final model by the analysis. Select this check box to use a single pooled estimate of variance that is computed by pooling variances across different studies for the shelf life estimation. When the check box is cleared, variances will not be pooled and the shelf life estimation will be based on separate variances computed from different studies. When studies with zero variance are requested to be included in the computations by clearing the Remove studies with zero variance check box, the Use pooled MSE for unequal int/unequal slope model check box should be selected.
Perform residual analysis of selected model
Select this check box to produce the following residual analyses:
- Predicted vs. studentized residual scatterplot
- Histogram of studentized residuals
- Line plot of studentized residuals in observation order
- Normal probability plot of raw residuals
- Normal probability plot of studentized residuals
The formula for the studentized residual for ith observation is as follows:
where,
εi = residual for ith observation
hii = ith diagonal element of X(XT X)-XT
MSE = Mean squared error of selected regression model
Interrupt analysis if at least one study fails data check
When this check box is selected, if any batch fails the input data checks, the analysis will halt and a message will be displayed listing which batches failed which data checks. The data checks are:
a. Invariance in the response variable
b. Insufficient time points (specified in the option for minimum number of time points per study)
c. Insufficient number of cases
Remove studies with zero variance
Select this check box to remove studies with zero variance. When the check box is cleared, studies with zero variance will be included in the computation and the Use pooled MSE for unequal int/unequal slope model check box should be selected, in case an unequal int/unequal slope model is encountered.
Generate predictions for the following time points
Enter a space-delimited set of time points for which you want to generate predictions based on the selected model. Confidence and prediction intervals are also computed.
Display date/time creation in subtitle of graphs
When selected, date and time creation is listed in the subtitle of graphs.