Hi,
I ran a coxph model and in the summary the independent variables are
reported highly significant. However, if I run anova() on that model, some
of the variables are reported to explain basically no deviance and other
variables which are reported insignificant in the model summary are reported
to explain some deviance here. Can somebody give advice how to make sense of
this output and interpret the results if the two outputs are somewhat
contradictory? Is it due to anova() including the variables sequentially? Do
I rely on the model or on the anova for interpretation?
Thanks much,
Daniel
Call:
coxph(formula = Surv(DurBeg, Closed) ~ SumInvested + numberofgainstocksheld
+
numberoflossstocksheld + TotalInvested + WorseThanMarket +
Loss + posilow + posihigh + negalow + negahigh + Loss:posilow +
Loss:posihigh + Loss:negalow + Loss:negahigh)
n=11232 (28 observations deleted due to missingness)
coef exp(coef) se(coef) z p
SumInvested -2.24e-05 1.000 1.12e-05 -1.999 4.6e-02
numberofgainstocksheld -5.37e-02 0.948 1.20e-02 -4.493 7.0e-06
numberoflossstocksheld -1.72e-01 0.842 1.22e-02 -14.032 0.0e+00
TotalInvested 3.18e-05 1.000 1.65e-05 1.919 5.5e-02
WorseThanMarket1 -9.83e-02 0.906 4.60e-02 -2.139 3.2e-02
Loss1 3.45e-01 1.412 9.67e-02 3.570 3.6e-04
posilow -6.72e-02 0.935 3.73e-02 -1.803 7.1e-02
posihigh 6.93e-02 1.072 3.62e-02 1.914 5.6e-02
negalow 1.87e-01 1.206 4.28e-02 4.367 1.3e-05
negahigh -2.42e-01 0.785 4.78e-02 -5.060 4.2e-07
Loss1:posilow 9.06e-03 1.009 5.25e-02 0.172 8.6e-01
Loss1:posihigh -1.04e-01 0.901 5.07e-02 -2.054 4.0e-02
Loss1:negalow -2.98e-01 0.742 6.10e-02 -4.891 1.0e-06
Loss1:negahigh 2.66e-01 1.305 6.70e-02 3.971 7.2e-05
exp(coef) exp(-coef) lower .95 upper .95
SumInvested 1.000 1.000 1.000 1.000
numberofgainstocksheld 0.948 1.055 0.926 0.970
numberoflossstocksheld 0.842 1.187 0.822 0.863
TotalInvested 1.000 1.000 1.000 1.000
WorseThanMarket1 0.906 1.103 0.828 0.992
Loss1 1.412 0.708 1.168 1.707
posilow 0.935 1.069 0.869 1.006
posihigh 1.072 0.933 0.998 1.151
negalow 1.206 0.829 1.109 1.311
negahigh 0.785 1.274 0.715 0.862
Loss1:posilow 1.009 0.991 0.910 1.118
Loss1:posihigh 0.901 1.110 0.816 0.995
Loss1:negalow 0.742 1.348 0.658 0.836
Loss1:negahigh 1.305 0.766 1.144 1.488
Rsquare= 0.035 (max possible= 0.967 )
Likelihood ratio test= 399 on 14 df, p=0
Wald test = 400 on 14 df, p=0
Score (logrank) test = 412 on 14 df, p=0
> anova(cluster4,test="Chisq")
Analysis of Deviance Table
Cox model: response is Surv(DurBeg, Closed)
Terms added sequentially (first to last)
Df Deviance Resid. Df Resid. Dev P(>|Chi|)
NULL 11232 38469
SumInvested 1 0 11231 38526 1
numberofgainstocksheld 1 55 11230 38471 1.032e-13
numberoflossstocksheld 1 271 11229 38200 7.454e-61
TotalInvested 1 4 11228 38196 3.964e-02
WorseThanMarket 1 7 11227 38189 8.439e-03
Loss 1 13 11226 38176 4.021e-04
posilow 1 66 11225 38110 4.306e-16
posihigh 1 1.502e-01 11224 38110 1
negalow 1 3.036e-01 11223 38110 1
negahigh 1 10 11222 38100 1.321e-03
Loss:posilow 1 1 11221 38099 3.385e-01
Loss:posihigh 1 2 11220 38096 1.208e-01
Loss:negalow 1 11 11219 38085 9.707e-04
Loss:negahigh 1 16 11218 38070 7.283e-05