91影视

All examples in the section involve not more than 20 samples for each bootstrap which is because of printing limitations. It is not convenient to use thousands of bootstrap samples.

When bootstrapping is done for real applications, at least 1000 samples should be used. Usually, statisticians use more than 10,000 bootstrap samples for real application.

The reason is to use so many samples is the more samples you use, the confidence interval is going to be accurate.

When only 20 bootstrap samples are used, there is a lot of variability at the lower and upper end of the confidence interval.

The confidence interval may be wider for more bootstrap samples hence the variability will be reduced which will give more accurate results.

For a real application, at least 1000 samples should be used for bootstrapping to get more accuracy.