Confidence Limits

Confidence Limits with no modifier usually refers to Confidence Limits on a Long Term Average. Confidence limits are two numbers (“upper and lower”) which form a claim that the average that would be obtained from a very large number of replicates lies between the two limits. Usually the claim is written at the 95% confidence level: “95% of the times that I make claims like this I am correct.” The computation of the confidence limits is done in a way that provides that 95% confidence level.

Confidence Limits (CL’s) can also be written to include the next single value observed, or to include 99% of future replicates (“tolerance limits”). STRATEGY for Windows also provides confidence limit for standard deviation and for b coefficients.