GEE (generalized gestimating equations)

[안재형]

그동안 gee()가 안되었었는데 방금 확인해보니 summary()없으면 error message가 뜨는데 summary를 넣으면 맞는 결과가 나오네요.

data(Rail, package="nlme")
library(gee)
summary(gee(travel~1, id=Rail, data=Rail, family=gaussian, corstr="exchangeable"))

Call:
gee(formula = travel ~ 1, id = Rail, data = Rail, family = gaussian,
    corstr = "exchangeable")

Summary of Residuals:
   Min     1Q Median     3Q    Max
-40.50 -16.25   0.00  18.50  33.50

Coefficients:
            Estimate Naive S.E.  Naive z Robust S.E. Robust z
(Intercept)     66.5   9.554026 6.960417    9.284844  7.16221

Estimated Scale Parameter:  559.0882
Number of Iterations:  1

Working Correlation
         [,1]     [,2]     [,3]
[1,] 1.000000 0.969383 0.969383
[2,] 0.969383 1.000000 0.969383
[3,] 0.969383 0.969383 1.000000

[강성찬]

잘 봤습니다. 그런데 이해를 하려면 책보고 공부좀 해야겠네요.^^

[안재형]

mixed models 하고 비슷합니다.