重复测量 - MIXED混合模型

在重复测量GLM中说到，随机效应那是被处理成了固定效应，而不是真正的随机效应。

1. GLM处理重复测量时的缺点

The assumption of compound symmetry required by the ‘univariate’ approach just discussed means that the correlation between a measurement at Time i and any other measurement within the same patient, say at Time j, is the same for all i and j.

假设复合对称，就是同一个patient，任意两个测量点之间的相关关系是相同的。这个假设很苛刻，尤其是有多个测量时间点时。

2. Mixed Model

上述问题，Mixed模型中可通过 TYPE = ，根据需求指定。

如果TYPE = UN: 两两比较，C42加上4个方差。因为这是假设任意两个协方差都不一致。

The R matrix is, therefore, block diagonal with 27 blocks, each block consisting of identical 4 x 4 unstructured matrices. The 10 parameters of these unstructured blocks make up the covariance parameters estimated by maximum likelihood

如果TYPE = CS: 就是2个参数。协方差全都相同（只有一个），只有一个方差，一个协方差。

第一块是第一个人有四条观测，第二块是第二个人有两条观测，第三块是第三个人有三个观测。

data arthr;

input vacgrp $ pat mo1 mo2 mo3;
datalines;
ACT 101 6 3 0
ACT 103 7 3 1
ACT 104 4 1 2
ACT 107 8 4 3
PBO 102 6 5 5
PBO 105 9 4 6
PBO 106 5 3 4
PBO 108 6 2 3
;

data discom; set arthr;
keep vacgrp pat visit score;
score = mo1; visit = 1; output;
score = mo2; visit = 2; output;
score = mo3; visit = 3; output;
run;

proc mixed data = discom;
class vacgrp pat visit;
*包含两个固定效应和一个交互效应;
model score = vacgrp visit vacgrp*visit;
*表明visits represent repeated measures within pat(vacgrp).;
repeated visit / subject=pat(vacgrp) type=un r rcorr;

*diff相当于两样本t检验,

slice相当于分组检验,by visit

如果都不加，相当于单样本t检验，均值不等于0得检验

;
lsmeans vacgrp visit vacgrp*visit / diff slice = visit;
estimate 'Month 3 Change from Baseline: ACT v. PBO'
vacgrp * visit -1 0 1 1 0 -1;
run;

repeated visit / subject=pat(vacgrp)

意思是visit代表pat() 中重复测量。visits represent repeated measures within pat(vacgrp)