[R] geeglm estimates and standard deviation are too large

kellyqiqi kellyqiqi at gmail.com
Thu Sep 29 16:12:46 CEST 2011


Hi,

I'm using geeglm function to account for the repeated measure. 

fit1<- geeglm( binary.outcome ~ age + race + gender + fever.yes.no,
data=mydata, id=ID, family=binomial, corstr="exchangeable")

summary(fit1)$coef gives too large estimates and standard deviation:

                                Estimate  Std.err    Wald Pr(>|W|)
(Intercept)                 3.07e+16 7.20e+14 1821.29  0.00000
age                           6.38e+13 2.22e+13      8.24  0.00409
RACEBlack                 1.48e+16 6.28e+14   555.35  0.00000
RACEOther               -1.84e+16 6.17e+14   887.78  0.00000
SEXFemale                1.84e+16 5.23e+14 1235.19  0.00000
FEVERYes                 -4.41e+15 4.74e+14   86.73  0.00000
FEVERUnknown          1.76e+16 1.60e+15  120.55  0.00000

compared to the estimates from the glm model:
                                Estimate Std. Error  z value Pr(>|z|)
(Intercept)               -0.65924   6.41e-01 -1.02875   0.3036
age                          0.00686   1.16e-02  0.59304   0.5532
RACEBlack                0.60687   4.13e-01  1.46900   0.1418
RACEOther              -1.18660   1.24e+00 -0.96054   0.3368
SEXFemale               0.61805   3.57e-01  1.73021   0.0836
FEVERYes               -0.96825   3.77e-01 -2.56554   0.0103
FEVERUnknown        0.39761   9.68e-01  0.41087   0.6812

I have 160 observations in my data, and 146 unique ID. Is that the problem?
Because I don't have "enough" repeated measures for each ID? 

Thank you very much 


--
View this message in context: http://r.789695.n4.nabble.com/geeglm-estimates-and-standard-deviation-are-too-large-tp3855902p3855902.html
Sent from the R help mailing list archive at Nabble.com.



More information about the R-help mailing list