Question

我构建了一个cox回归：

Call:
coxph(formula = Surv(`infection time`, infection) ~ Treatment + 
    Gender + Race + Surface + head + buttock + trunk + up.leg + 
    low.leg + resp.tract, data = BurnData)

整体测试的p值（似然，Wald和得分）仅对似然比检验有意义。这些测试评估$ H_0 $：所有$ \ beta $ = 0.测试统计数据不在协议中，因此无法明确拒绝原假设。总的来说，该模型并不适合。

我想了解协变量治疗，特别是我想知道两个治疗组在感染之前是否存在显着差异？从上面的输出我可以说协变量是重要的，但这是不一样的吗？

我尝试过strata命令：

coxph(formula = Surv(`infection time`, infection) ~ strata(Treatment) + 
    Gender + Race + Surface + head + buttock + trunk + up.leg + 
    low.leg + resp.tract, data = BurnData)

  n= 154, number of events= 48 

                coef exp(coef)  se(coef)      z Pr(>|z|)  
Gender     -0.649207  0.522460  0.396943 -1.636   0.1019  
Race        1.997690  7.372010  1.016640  1.965   0.0494 *
Surface     0.003237  1.003243  0.009704  0.334   0.7387  
head       -0.082786  0.920548  0.344487 -0.240   0.8101  
buttock     0.560418  1.751404  0.421308  1.330   0.1835  
trunk       0.239397  1.270483  0.494814  0.484   0.6285  
up.leg     -0.432832  0.648669  0.375035 -1.154   0.2485  
low.leg    -0.164111  0.848648  0.373298 -0.440   0.6602  
resp.tract  0.077946  1.081065  0.351205  0.222   0.8244  
---
Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

           exp(coef) exp(-coef) lower .95 upper .95
Gender        0.5225     1.9140    0.2400     1.137
Race          7.3720     0.1356    1.0051    54.069
Surface       1.0032     0.9968    0.9843     1.023
head          0.9205     1.0863    0.4686     1.808
buttock       1.7514     0.5710    0.7670     3.999
trunk         1.2705     0.7871    0.4817     3.351
up.leg        0.6487     1.5416    0.3110     1.353
low.leg       0.8486     1.1783    0.4083     1.764
resp.tract    1.0811     0.9250    0.5431     2.152

Concordance= 0.67  (se = 0.066 )
Rsquare= 0.098   (max possible= 0.909 )
Likelihood ratio test= 15.89  on 9 df,   p=0.06912
Wald test            = 10.44  on 9 df,   p=0.3157
Score (logrank) test = 12.71  on 9 df,   p=0.1761

但是，现在治疗变量根本不存在。如果两个不同的组对感染时间的影响有显着差异，我该如何比较？

我也试过了anova comparison。以下是有效的方法吗？

fit_trt0 <- coxph(Surv(`infection time`,infection) ~ (Treatment==0) + Gender + Race + Surface + 
                   head + buttock + trunk + up.leg + low.leg + resp.tract, data=BurnData)
summary(fit_trt0)

fit_trt1 <- coxph(Surv(`infection time`,infection) ~ (Treatment==1) + Gender + Race + Surface + 
                   head + buttock + trunk + up.leg + low.leg + resp.tract, data=BurnData)
summary(fit_trt1)

anova(fit_trt0,fit_trt1)

比较R - anova中cox回归的组间差异？

0 个答案: