对前三种因素进行时间序列分析CPI程序：dataa;inputyearCPI@@;cards;201102104.9201101104.9201012104.6201011105.1201010104.4201009103.6201008103.5201007103.3201006102.9201005103.1201004102.8201003102.4201002102.7201001101.5200912101.9200911100.620091099.520090999.220090898.820090798.220090698.320090598.620090498.520090398.820090298.4200901101200812101.2200811102.4200810104200809104.6200808104.9200807106.3200806107.1200805107.7200804108.5200803108.3200802108.7200801107.1;procgplot;plotCPI*year=1;symbol1v=diamondi=joinc=red;run;quit;procarimadata=a;identifyvar=CPI;estimatep=1method=ml;Run;图一图二样本自相关图延迟3阶后，自相关系数落入2倍范围以内，而且自相关系数向零递减的速度非常快，可以认为该序列非常平稳。图三因此，我们可以考虑用AR（1）模型来拟合该观察值序列。图四图四残差白噪声检验显示只有延迟6阶LB检验统计量的P值明显大于0.05，所以该AR（1）模型无效。GDP:程序如下：dataA;inputyearGDP@@;cards;201102107.2201101106.6201012105.9201011106.1201010105201009104.3201008104.3201007104.8201006106.4201005107.1201004106.8201003105.9201002105.4201001104.3200912101.720091197.920091094.22009099320090892.120090791.820090692.220090592.820090493.42009039420090295.520090196.720081298.9200811102200810106.6200809109.1200808110.1200807110200806108.8200805108.2200804108.1200803108200802106.6200801106.1200712105.4200711104.6200710103.2200709102.7200708102.6200707102.4200706102.5200705102.8200704102.9200703102.7200702102.6200701103.3200612103.1200611102.8200610102.9200609103.45200608103.4200607103.6200606103.5200605102.4200604101.9200603102.5200602103200601103.1;procgplot;plotGDP*year=1;symbol1v=diamondi=joinc=red;procarimadata=A;identifyvar=GDP;estimatep=1method=ml;Run;quit;图一样本自相关图延迟3阶之后，自相关系数都落入2倍标准差范围以内，而且自相关系数向零衰减的速度非常快，延迟8阶之后自相关系数即在零值附近波动。这是一个非常典型的短期相关的样本自相关图。由时序图和样本自相关图的性质，可以认为该序列平稳。考虑用AR（1）模型來拟合该观察序列图二图三在延迟18阶、24阶LB检验统计量的P值均显著大于0.05，所以该模型有效。图四由图四即得该模型为：GDP=103.72899+§／(1+0.95489)