6准备金评估的GLM演示教学.pptx

准备金评估的GLM 孟生旺中国人民大学统计学院/mengshwGLM模型的结构分布假设线性预测项连接函数GLM模型的检验和比较关于解释变量：是否显著？方差分析，P值是否分段？是否线性？平滑函数是否存在交互效应？关于分布假设：残差分析：QQ图，蠕虫图。流量三角形的数据有限，分布假设的影响显著关于连接函数？模型比较：AIC = - 2l + 2*参数个数增量赔款的流量三角形格式（Taylor，1983）事故年进展年123456789101357,848766,940610,542482,940527,326574,398146,342139,950227,22967,9482352,118884,021933,8941,183,289445,745320,996527,804266,172425,046?3290,5071,001,799926,2191,016,654750,816146,923495,992280,405??4310,6081,108,250776,1891,562,400272,482352,053206,286???5443,160693,190991,983769,488504,851470,639????6396,132937,085847,498805,037705,960?????7440,832847,6311,131,3981,063,269??????8359,4801,061,6481,443,370???????9376,686986,608????????10344,014?????????增量赔款的数据框格式事故年进展年增量赔款11357848127669401361054214482940155273261657439817146342181399501922722911067948213521182288402123933894241183289254457452632099627527804282661722942504631290507………描述性统计分析第1个事故年的进展趋势比较特殊？均值和方差Min.1st Qu.MedianMean3rd Qu.Max.679503521005273006247009051001562000方差 = 122188299213方差/均值 = 195597泊松分布拟合：图中泊松分布的概率被缩小为实际值的1%伽马分布拟合描述性分析的初步结论增量赔款右偏，方差远远大于均值，伽马拟合较好第1个事故年的进展模式与其它事故年不同建模分布假设：泊松（链梯法）伽马解释变量：事故年、进展年、日历年形式：离散、连续连接函数：对数泊松回归增量赔款 ~ 泊松分布log(增量赔款) = 事故年 + 进展年链梯法（使用累积赔款数据）12345678910准备124,7881,735,3302,218,2702,745,5963,319,9943,466,3363,606,2863,833,5153,901,463236,1392,170,0333,353,3223,799,0674,120,0634,647,8674,914,0395,339,0855,433,71994,63432905071,292,3062,218,5253,235,1793,985,9954,132,9184,628,9104,909,315　5,378,826469,51143106081,418,8582,195,0473,757,4474,029,9294,381,9824,588,268　　5,297,906709,63854431601,136,3502,128,3332,897,8213,402,6723,873,311　　　4,858,200984,88963961321,333,2172,180,7152,985,7523,691,712　　　　5,111,1711,419,45974408321,288,4632,419,8613,483,130　　　　　5,660,7712,177,64183594801,421,1282,864,498　　　　　　6,784,7993,920,30193766861,363,294　　　　　　　5,642,2664,278,97210344014　　　　　　　　4,969,8254,625,811进展因子37473331.4574131.1738521.1038241.0862691.0538741.0765551累积进展因子14.4465774.1387012.3685821.6251961.3844991.2542761.1546641.095637