matlab线性回归s和2,数据回归分析和拟合的matlab实现2.doc

数据回归分析和拟合的matlab实现2.doc

数据回归分析和拟合的MATLAB实现目录一、多元线性回归二、多项式回归一元多项式POLYFIT或者POLYTOOL多元二项式RSTOOL或者RSMDEMO三、非线性回归四、逐步回归一、多元线性回归多元线性回归1、BREGRESSY,X确定回归系数的点估计值2、B,BINT,R,RINT,STATSREGRESSY,X,ALPHA求回归系数的点估计和区间估计、并检验回归模型①BINT表示回归系数的区间估计②R表示残差③RINT表示置信区间④STATS表示用于检验回归模型的统计量,有三个数值相关系数R2、F值、与F对应的概率P说明相关系数R2越接近1，说明回归方程越显著；F越大，说明回归方程越显著；与F对应的概率PX143145146147149150153154155156157158159160162164 XONES16,1XY8885889192939395969897969899100102 2回归分析及检验B,BINT,R,RINT,STATSREGRESSY,XB16073007194BINT337071156120604708340R12056323310952413282088951170209879029270573418540013471584703040002340462100992RINT1240736520506221404035894168451289539459185193630915552389553771317955254733132822471339390754044621268142950842188104943071024630276612719331133218922464026624STATS0928218095310000017437运行结果解读如下参数回归结果为对应的置信区间分别为337017,15612和06047,0834R209282越接近于1，回归效果越显著，F1809531，P00000，由PT1/301/3014/30S11861567206026693371419351136149729085449908113771295414648P,SPOLYFITT,S,2P489294665889691329SR3X3DOUBLEDF11NORMR01157解法二化为多元线性回归T1/301/3014/30S11861567206026693371419351136149729085449908113771295414648TONES14,1T T2 B,BINT,R,RINT,STATSREGRESSS ,TB913296588964892946BINT906149204465231666547648801464905747R0012900302001480073200040004740016500078003630022200046000590023700411RINT00697004390095600352008760058000182012830070900789001920113900894005630081300658010620033500955005110070400796007930067500904004290008800910STATS10E0070000010378000000故回归模型为预测及作图YPOLYCONFP,T,S；PLOTT,S, K ,T,Y, R 1、多元二项式回归MATLAB命令RSTOOLX,Y, MODEL ,ALPHA输入参数说明XNM矩阵；YN维列向量；ALPHA显著性水平缺省时为005；MODE由下列4个模型中选择1个用字符串输入,缺省时为线性模型2、实例演示说明设某商品的需求量与消费者的平均收入、商品价格的统计数据如下,建立回归模型,预测平均收入为1000、价格为6时的商品需求量需求量10075807050659010011060收入10006001200500300400130011001300300价格5766875439解法一选择纯二次模型直接用多元二项式回归如下X110006001200500300400130011001300300X25766875439Y10075807050659010011060 XX1 X2 RSTOOLX,Y, PUREQUADRATIC 二、多项式回归一元多项式回归1、一元多项式回归函数1P,SPOLYFITX,Y,M确定多项式系数的MATLAB命令说明XX1,X2,,XN,YY1,Y2,,YN；PA1,A2,,AM1是多项式YA1XMA2XM1AMXAM1的系数；S是一个矩阵,用来估计预测误差2POLYTOOLX,Y,M调用多项式回归GUI界面，参数意义同POLYFIT2、预测和预测误差估计1YPOLYVALP,X求POLYFIT所得的回归多项式在X处的预测值Y2Y,DELTAPOLYCONFP,X,S,ALPHA求POLYFIT所得的回归多项式在X处的预测值Y及预测值的显著性为1ALPHA的置信区间Y±DELTA，ALPHA缺省时为053、实例演示说明观测物体降落的距离S与时间T的关系，得到数据如下表，求S的表达式即回归方程SABTCT2TS1/302/303/304/305/306/307/30SCM1186156720602669337141935113TS8/309/3010/3011/3012/3013/3014/30SCM6149729085449908113771295414648解法一直接作二次多项式回归12T1/301/3014/303S118615672060266933714193511361497290854499081137712954146484P,SPOLYFITT,S,256P7848929466588969132991011S1213R3X3DOUBLE14DF1115NORMR01157复制代码故回归模型为解法二化为多元线性回归12T1/301/3014/303S118615672060266933714193511361497290854499081137712954146484TONES14,1T T2 5B,BINT,R,RINT,STATSREGRESSS ,T67B899132910658896114892946121314BINT15169061492044176523166654761848801464905747192021R2223001292400302250014826007322700040280047429001653000078310036332002223300046340005935002373600411373839RINT40410069700439420095600352430087600580440018201283450070900789460019201139470089400563480081300658490106200335500095500511510070400796520079300675530090400429540008800910555657S