Research Article

Optimal Bandwidth Selection for Kernel Density Functionals Estimation

Table 1

Optimal bandwidth selection in estimation of with comparison to classical bandwidth selection methods.

Bandwidth selection for estimation of
Proposed method Classical bandwidth for KDE
100 100 100 100 100 100

Normal
 Mean .197 .154 .236 .171 .221 .163 .302 .270 .359 .298 .300 .268
 1st quantile .182 .144 .205 .156 .200 .153 .278 .254 .307 .258 .265 .243
 Median .197 .154 .233 .170 .227 .167 .301 .270 .376 .317 .300 .268
 3rd quantile .214 .163 .265 .185 .254 .183 .327 .287 .431 .355 .336 .293
Cauchy
 Mean .325 .240 .320 .222 .290 .205 .497 .422 .440 .331 .419 .338
 1st quantile .273 .214 .264 .198 .230 .174 .417 .377 .330 .265 .348 .310
 Median .316 .236 .312 .219 .295 .207 .484 .415 .455 .353 .416 .342
 3rd quantile .365 .265 .368 .245 .355 .238 .557 .465 .554 .415 .478 .387
Pareto
 Mean .442 .332 .263 .178 .145 .098 .676 .583 .207 .143 .374 .284
 1st quantile .329 .267 .210 .151 .103 .078 .502 .469 .140 .106 .298 .241
 Median .422 .320 .251 .174 .137 .091 .646 .563 .200 .131 .359 .278
 3rd quantile .519 .386 .300 .202 .174 .109 .794 .678 .260 .173 .428 .327
Mix-Normal I
 Mean .378 .289 .389 .279 .349 .253 .578 .507 .522 .413 .495 .417
 1st Quantile .360 .279 .342 .252 .282 .211 .551 .490 .389 .320 .438 .376
 Median .377 .288 .382 .273 .362 .259 .577 .506 .510 .403 .489 .407
 3rd Quantile .397 .298 .431 .309 .428 .304 .607 .524 .658 .510 .544 .460
Mix-Normal II
 Mean .342 .263 .364 .255 .322 .227 .522 .461 .492 .372 .467 .389
 1st Quantile .321 .251 .318 .229 .254 .188 .490 .441 .339 .279 .413 .347
 Median .342 .262 .361 .254 .345 .231 .523 .461 .509 .355 .464 .387
 3rd Quantile .364 .274 .403 .278 .397 .275 .556 .481 .642 .461 .519 .425

ROT bandwidth for estimation of given by (12); 2DPI bandwidth for estimation of with pilot bandwidth ; 3DPI bandwidth for estimation of with pilot bandwidth ; 4ROT bandwidth for density estimation proposed in Silverman [7]; 5LSCV bandwidth for density estimation proposed in Bowman [8]; 6DPI bandwidth for density estimation proposed in Sheather and Jones [9].