Question

我有一些SAS代码运行三个不同的GLMM，我需要在R中运行相同的。任何人都可以帮我翻译吗？我之前从未与SAS合作过。我一直在使用＆＃39; glmer.nb＆＃39;来自lme4，但我愿意在必要时改变。 SAS代码如下。

libname dataset 'H:\ReefSAS';

PROC IMPORT DATAFILE = "H:\ReefSAS\reefdatasas.xls"
OUT = reefdata DBMS = xls REPLACE;
GETNAMES = yes;
RUN;

/*] "date"  "reefID"  "station"  "densClass" 
"easting"  "northing"  "altitude"  "density"  */

/* three different spatial structures  */

proc mixed data=reefdata;
  class date station densClass;
  model density =station densClass;
  Random reefID;
  repeated date/ TYPE=SP(EXPGA)(easting northing) subject=reefID; 
run;

proc mixed data=reefdata;
  class date station densClass;
  model density =station densClass;
  Random reefID;
  repeated date/ TYPE=SP(GAUGA)(easting northing) subject=reefID; 
run;

proc mixed data=reefdata;
  class date station densClass;
  model density =station densClass;
  Random reefID;
  repeated date/ TYPE=SP(SPHGA)(easting northing) subject=reefID; 
run;

编辑：我当前的代码有三个选项，一个工作（lme :: nlme），两个不工作（glmer.nb :: lme4和nlme :: nlme）。说实话，我不确定我是否正确地翻译了SAS的陈述。我还想指出，此刻的配方是线性的，但我的想法是，一旦我开始运行，就以非线性方式用其他变量的组合来建立固定效果。我到目前为止的代码：

library(lme4) # works, but it does not account for spatial auto-correlation
glmer.nb(density~date+densClass+1|reefID, data=reefdata)

library(nlme) # does not run, error below
nlme(model=density~alpha*date + beta*densClass, data=reefdata,
         fixed=alpha+beta~date*densClass,
         random=density~1|reefID,
         start=list(fixed=0.05, alpha=0.05, beta=0.05),
         correlation=corExp(form=~easting+northing |reefID))
# Error in nlme.formula(model = densAVG ~ alpha * dateLvls + beta * densClass, : starting values for the 'fixed' component are not the correct length

library(nlme) # works but does not incorporate spatial auto-correlation
lme(density~date+densClass, random=~1|reefID, data=reefdata)

将具有空间自相关协方差结构的GLMM的SAS代码转换为R.

0 个答案: