In statistical analyses estimation of unknown parameter(s) and hypotheses testing are common practices. But all estimators or tests don't have equally good statistical properties. Hence the users are required selecting the best estimator or test from the available alternatives. Both parameter estimation and hypothesis testing require sample data. A natural expectation is that inclusion of relevant non-sample prior information in the estimation and testing process would result in improving the quality of statistical decisions. This presentation demonstrates that the preliminary test and shrinkage estimators often have better statistical properties than the commonly used exclusive sample data based estimators. The notion of inclusion of non-sample prior information is extended to the testing regime. It is evident that the use of non-sample prior information increases the power of detecting any false null hypothesis. Also the chance of rejecting a correct null hypothesis, in other words the size of the test, is minimized by testing one parameter after pre-testing another parameter.