Row B: Tests sep. intercep, com slope vs. com intercep, com slope

The slopes can be considered equal for all of the batches. You must verify whether the intercepts can be considered equal as well. The null hypothesis is that the lines you draw for each of the batches cross the Y axis at a similar point that they are basically the same, unless you have evidence to reject this hypothesis.

You no longer need to include the interaction term in this model. Hence, Time and Batch are included.

If you only have Time in the model, you have a single regression line. All of the points contribute to the one line, regardless of the batch. There is one common slope and one common intercept.

Batch Term

Batch term is the term that determines whether additional variance is explained by having intercepts rather than a common intercept.

The Batch term has a p-value of 0.001361 which is less than the chosen value of .25. Since the Batch term is significant, you must keep the lines separate. They are too different to combine. Select the Sep. Intercept, Common Slope model.

In a case where the p-value is not significant (below the chosen value of 0.25), you can pool all of the batches together into a single regression line. Select the Com. Intercept, Com. Slope model.

Stability Graph

A Stability graph with the chosen model is shown later in this section. The regression lines are drawn for each batch. The lines are parallel to one another (that is, have the same slope), but intersect the zero time line at different points. The red horizontal line represents the upper acceptable limit for this parameter value. The shelf life is determined by default by the time value where the first 95% confidence limit crosses the defined upper limit. In the following graph, the MNO batch line's confidence limit (pink dotted line) determines the shelf life of 29.91 months.