[R-sig-Epi] Cox Hazard analysis using unequal number of data

[Adriana Humanes]

I'm doing a survival analysis of a temperature treatment on 3 different
sexual cross types of corals. I have a data set of 628 observations with
unequal number of samples (see table below). I would like to know the power
if the low number of samples in the HL treatment is affecting my results in
the cox hazard regression.

Thanks in advance,

N Observed Expected
Treatment_Cross=Control_HighHigh 106
Treatment_Cross=Control_HighLow 69
Treatment_Cross=Control_LowLow 173
Treatment_Cross=Temperature_HighHigh 99
Treatment_Cross=Temperature_HighLow 59
Treatment_Cross=Temperature_LowLow 117


coxph(formula = Surv(Age, Status) ~ Treatment * Cross_type, data =
Recruits_survival)

n= 628, number of events= 225

coef exp(coef) se(coef) z Pr(>|z|)

TreatmentT 0.12375 1.13173 0.21321 0.580 0.56165
Cross_typeHL -0.78216 0.45741 0.29906 -2.615 0.00891 **
Cross_typeLL -0.09413 0.91017 0.19348 -0.486 0.62662
TreatmentT:Cross_typeHL 0.36443 1.43969 0.40265 0.905 0.36543
TreatmentT:Cross_typeLL -0.48355 0.61659 0.29932 -1.615 0.10620
---
Signif. codes: 0 ‘’ 0.001 ‘’ 0.01 ‘’ 0.05 ‘.’ 0.1 ‘ ’ 1
exp(coef) exp(-coef) lower .95 upper .95
TreatmentT 1.1317 0.8836 0.7452 1.719
Cross_typeHL 0.4574 2.1862 0.2545 0.822
Cross_typeLL 0.9102 1.0987 0.6229 1.330
TreatmentT:Cross_typeHL 1.4397 0.6946 0.6539 3.170
TreatmentT:Cross_typeLL 0.6166 1.6218 0.3429 1.109

Concordance= 0.582 (se = 0.02 )

Likelihood ratio test= 15.27 on 5 df, p=0.009

Wald test = 14.03 on 5 df, p=0.02

Score (logrank) test = 14.51 on 5 df, p=0.01

Adriana Humanes

www.adrianahumanes.com

www.coralassistlab.org


Postdoctoral Research Associate

Ridley Building 2
Newcastle University
Claremont Road
Newcastle upon Tyne NE1 7RU
Telephone: +44 (0)191 208 6441 <+44%20191%20208%206441>

	[[alternative HTML version deleted]]