Example of a prediction for a stability studymain topic interpreting results session command see also
Example of a prediction for a stability studyIn Example of a stability study, you determine that the shelf life of a new medication is 54.79 months. The shelf life for this analysis is defined as the time at which you can no longer be 95% confident that the strength of the worst batch is 90% of the intended strength. You want to predict the mean strength for the best batch and the worst batch at 54.79 months.
You do not need to re-analyze the stability study model. The worksheet contains the model for the prediction.
1 Open the worksheet SHELFLIFE2.MTW.
2 Choose Stat > Regression > Stability Study > Predict.
3 In Response, choose Drug%.
4 In the second drop-down list, choose Enter individual values.
5 In the variables table, enter the setting for each variable as shown below.
|
Month |
Batch |
|
54.79 |
1 |
|
54.79 |
2 |
6 Click OK.
Session Window Output
Prediction for Drug%
Regression Equation
Batch 1 Drug% = 99.853 - 0.090918 Month
2 Drug% = 100.15 - 0.16047 Month
Variable Setting Month 54.79 Batch 1
Fit SE Fit 95% CI 95% PI 94.8716 0.801867 (93.2340, 96.5092) (92.8324, 96.9108) XX
Variable Setting Month 54.79 Batch 2
Fit SE Fit 95% CI 95% PI 91.3609 0.801867 (89.7233, 92.9986) (89.3217, 93.4001) XX
XX denotes an extremely unusual point relative to predictor levels used to fit the model. |
For the pill data, the predicted values from Batch 1 are calculated using the formula, Drug% = 99.853 - 0.090918 Month. The slope for Batch 1, 0.090918, is less steep than the slope for Batch 2, 0.16047, which indicates that the strength of Batch 1 decreases more slowly than the strength of Batch 2.
The batch that has the smallest decrease in drug strength over time is Batch 1. For Batch 1, the predicted drug strength at 54.79 months is approximately 94.87%. The predicted drug strength for Batch 2 is approximately 91.36%.
The XX next to each row indicates that the original data do not include the variable settings that you want to predict. Your predictions are for 54 months, but the oldest samples that you test are only 48 months old. Be cautious when you extend predictions beyond the original data because you cannot be certain that the relationships remain consistent. For example, the pills could loose strength more rapidly after 48 months. Only further testing with older samples can confirm that your shelf life estimate is accurate.
The 95% confidence interval (CIs) indicate that you can estimate with 95% confidence that the mean drug concentration for Batch 1 at 54.79 months is between approximately 93.23% and 96.51%.
The 95% prediction interval (PI) indicates that you can estimate with 95% confidence that the predicted strength for a single pill from Batch 1 at 54.79 months is between approximately 92.83% and 96.91%.