Products Reports: Dunnett's Analysis For Shelf-life Testing Data

Overview

The recommended analysis for shelf-life testing data is the Dunnett's test. Dunnett's test compares a (blind) control from the first test (time point 0) against each test sample from to subsequent tests (time points).

This workflow will help you select specific analysis options when generating the PowerPoint Report in the Products library to ensure proper Dunnett's analysis. The same steps can be applied to other report types, such as Summary Report, Top Box Report, and raw data export.

Before You Begin

Please be sure to follow the proper test setup workflow for shelf-life testing.
This analysis workflow is applicable to real-time shelf-life time point testing data and accelerated shelf-life testing data collected in multiple tests.
For analysis purposes it is very important that the Sample type of your time point 0 (first test's) control sample is set to Control. The steps below can be performed even after data is collected, in case it was not done before data collection.

If you already set your test to Complete, undo complete to be able to follow the remaining steps.

In the Samples & design tab of your time point 0 test, click Edit samples.

Locate the sample you wish to mark as a Control.

Click on the Sample type dropdown list for that sample and select Control .

Click Save to save the changes.

Set the test to Complete.

Only one sample should be set as a Control sample type for Dunnett's analysis to run successfully.

Generate The PowerPoint Report

From the black ribbon at the top, select Libraries > Products > Reports, and follow the steps below:

Under the 1. Select report type, select PowerPoint report .

This is where you can select a different report type if you wish to do so. Please be sure to review all the options for different reports to select what meets your analysis needs.
Under the 2. Products, click Select products:

Select the product group where your product resides.
Locate and select the product you wish to run analysis across tests on.
Click Use selected product(s).

Under the 3. Test criteria, click Edit criteria:

By default, Completion date criterion is set to between a month ago and tomorrow. Modify the date range as needed if you want to search for tests based on the Completion date.

If you wish to change the criterion, in the first dropdown list, click Completion date and select a different criterion that better suits your searching needs.
Every criterion will have its own set of operators. Review carefully and select the options that best suit your searching needs.
If you wish to add additional criteria to narrow down your search, click +Add criteria and set it up as necessary. Please note that by adding multiple criteria tests that meet all of the criteria will be included in the analysis. Tests that do not meet all of the criteria will be ignored.
Click Save criteria.

Under the 4. Tests, you will see how many tests met your test criteria specified in the above step. If you wish to review and further filter the tests, click Select tests. Deselect tests that you wish to exclude from the analysis and click Use selected test.

Maximum number of samples the Dunnett's test can run on is 17. If you have more than 17 samples, run the report two times. Either filter out the tests at this stage, or filter out some of the samples in the next step (5), making sure to include the control sample in both reports. By doing this you will still get the proper analysis because Dunnett's test compares each individual sample to the (blind) control.
Under the 5. Filters, click Edit filters to filter data, if so desired, using one, some, or all of the below filtering options:

Sample sets. You can filter out sample sets for individual tests selected in the list.
Samples. Samples from all tests are displayed in the tab. The test name they belong to will be displayed in the screen. Filter samples out as necessary.

Maximum number of samples the Dunnett's test can run on is 17. If you have more than 17 samples, run the report two times. Either filter out the tests at the previous step (4), or filter out some of the samples in this step, making sure to include the control sample in both reports. By doing this you will still get the proper analysis because Dunnett's test compares each individual sample to the (blind) control.
Reps. If any of the tests had reps defined in them, you will be able to filter reps out.

Under the 6. Options, make the following selections:
1. Under the Sample grouping options, select Treat each sample as unique. This will ensure that even if you accidentally forgot to update the sample name for one of the time points that the analysis will treat the samples as unique rather than group data for samples with the same name, for example.
  
  For report clarity you might want to ensure that the sample names are correct before running the reports. You can update sample names even when data is collected, but not if the test is set to Complete. If you already set the test to Complete and need to make changes to the sample names, undo complete, make the change, set the test back to Complete and run the analysis.
2. Review the Report format and details options and make selections as needed.
3. Under the Analysis options from defaults, click Change advanced analysis options.
4. Select Dunnett's test at the significance level of your choice and select the ANOVA model that meets your company standards.
Under the 7. Questions, click Choose questions, review and select the questions you wish to include in the analysis.
Under the 8. Export, the only option available for the PowerPoint Report is .pptx. Other report types will have different options to choose from.
Update the name for your report and click Create my report. Download and review the report.

Maximum number of samples the Dunnett's test can run on is 17. If you have more than 17 samples, run the report two times. Either filter out the tests at the step 4, or filter out some of the samples in the step 5, making sure to include the control sample in both reports.By doing this you will still get the proper analysis because Dunnett's test compares each individual sample to the (blind) control.

If your tests have an unequal number of responses, the Dunnett’s test will be run on a per pair basis.

This means that when you have an equal number of responses for each sample, there will be one Dunnett’s t value calculated and applied across all samples.

When you have an unequal number of responses for each sample, Dunnett’s t value will be calculated for each pair (and the column will not appear in the summary report because it is not the same across all samples). For example, control vs. sample 2 will use a different t value than control vs. sample 3.