Inferences of small samples - Brief points (Applied Econometrics)

 □ Methods for the small sample inference

Monte Carlo methods and Bootstrap method.

▪ Monte Carlo Method

This method helps in the choice between different estimators and test statistics.

▪ Bootstrap Method

This method helps in making small sample inference on the chosen estimator and test statistic.

There are 2 bootstrap data generation methods: the residual based method and thr direct method of bootstrapping the data.

 There are different methods for constructing confidence intervals - the percentile method and the percentile t method and its also that the latter method is the preferred choice.

■ Is the defective Bootstrap method still better than asymptotic inference ? 

Actually, there are many instances where defective bootstrap methods have been used. There are several examples in several literatures where the defective Bootstrap method still is not better than asymptotic inference. 

One of the examples, is in the unit root. However, when no asymptotic inference is available, its better to use bootstrap method. Also, when a correct bootstrap method is complicated, and not feasable, a theoritically imperfect bootstrap method might still improve asymptotic inference.

Thus, unless proven else, some bootstrap may be are better than no bootstrap.

Thank you

Note: Details will be discussed in upcoming blogs.

Aditya Pokhrel

MBA, MA Economics, MPA