Example - Nonrandomized Testing Order

The attached files show the impact on test results of not randomizing the order of testing.

• "ASTM Test - static load vs time (non-randomized trials).stw"

• "Example of the effect of not randomizing test order.doc"

The test method involved a static load on a product that was attached to a steel plate via the barrier (adhesive) material. The time to failure was recorded for each device.

The static load was applied using a piece of PVC tubing that was filled with lead shot. Loads of 14.25 lbs., 14.5 lbs., 14.75 lbs., and 15.0 lbs. were applied. Each load had 12 specimens tested.

The unfortunate result:

Ideally, the application of load should have been randomized (if not fully randomized, then at least partly randomized with a blocking variable). Instead, all specimens for a particular load were tested before testing the next set of specimens at a different load. This caused an unexplainable difference for the group tested at 14.5 lbs, which compromised the ability to draw conclusions from the study.

How could this have been prevented?

This could have been prevented by randomizing the order of testing of the loads.