Statistically maintaining initial value

[kernalms]

I am new to the post. I need help with a experimental problem. I am testing to see if radiation effects the emission from luminophores on sample substrates. I have very limited number of samples to test. My first problem was a measurement accuracy issue. The emission measurement device is notoriously sensitive to the slightest changes. In order to reduce the variation between measures, I take 16 measurements of each sample at each test point. Then I average the 16 measurements. This method has reduced my error by square root of 16 thus much more accurate results (usually ~2% error which is very accurate for my case). After implementing this solution, my next issue is to "explain" if my sample will maintain its initial value after radiation exposure. In other words, the test sample does not degrade with exposure. I have only been able to test a few samples due to constraints but feel I have accurate data at each test point.

My question: is there a method to show that the sample does not decrease below the initial value, thus it does not degrade with increased radiation exposure?

I have tried to use R^2 to show how it can predict the data results but most of my test points are still jumping enough that the R^2 value is low. I was thinking of comparing the 2 sigma guassing curves from the initial value and the last test point to show that the last test point has a high probability to be at or above the initial curve. Or using a Z-test to compare how far from an average that the test points deviate. Assuming the test points change minimally from the average (Z<2), that would show that the values are maintaining close to the average, thus not degrading. Any help in the method needed or even a topic to research to better explain this maintenance is greatly appreciated. Thank you.