SEVERITY ESTIMATION IN THE FIRE INSURANCE THROUGH SEMIPARAMETRIC BOOTSTRAP
Along with the development of information, science and technology, there is a resampling method that is quite popular to be developed, namely bootstrapping. Bootstrap estimates asymptotically against its original value (observation). Thus, the greater the bootstrap replication, the resampel distribu...
Saved in:
| Main Author: | |
|---|---|
| Format: | Theses |
| Language: | Indonesia |
| Online Access: | https://digilib.itb.ac.id/gdl/view/49699 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Institut Teknologi Bandung |
| Language: | Indonesia |
| Summary: | Along with the development of information, science and technology, there is a resampling method that is quite popular to be developed, namely bootstrapping. Bootstrap estimates asymptotically against its original value (observation). Thus, the greater the bootstrap replication, the resampel distribution will be normally distributed. This indicates that the bootstrap estimation gives better results. In this case study, the fire insurance data classified into two different claim criterion, namely Limited and Guaranteed and All Claim but Limited. The data processing show that the severity on fire insurance data for those criterion follow Weibull(?????,?????) distribution with different parameter estimators. ???? represents the scale parameter and ???? represents the shape parameter. Based on the goodness-of-fit test by using Kolmogorov-Smirnov test, the scale and shape parameter estimators on the Limited and Guaranteed criteria are 1,5708×108 and 0,682 respectively. Meanwhile, the scale and shape parameter estimators on the All Claim but Limited criteria are 1,0046 ×108 and 0,53873 respectively. However, there is no guarantee that the data come from a certain distribution. So that, the semiparametric bootstrap method suitable in estimating the severity. Based on the semiparametric bootstrap, the optimum bootstrap replication for bootstrap estimation can also be analyzed. Bootstrap estimators of mean and variance tend to converge quicker than for skewness and kurtosis. |
|---|