Comparison of The Efficiency of Hypotheses Tests and Confidence Intervals in the Inference Process. Study on Means

Abstract

Relevant studies ensure that when you need to infer about a parameter, confidence intervals are more appropriate alternatives than hypothesis testing. This study measures the efficiency of inference of these two methodologies, applied on contrasts of means, comparing the probability of an error and a success in each one of them ─These probabilities are estimated through a stochastic Monte Carlo simulation process. Taking into account this criterium, it was found that both inferential processes are equally effective, for this reason, it cannot be guaranteed that one procedure is better than another. Confidence intervals provide more information about the measurement of significant differences between the parameters compared, but this should not be confused with greater efficiency. The problem is not hypothesis testing models, but rather the approach they traditionally present. To overcome this problem and get than methodology based on P value provides the same inferential information as the confidence intervals, it is recommended to use an approach of hypothesis testing based on equivalence.