[R] svyglm and sandwich estimator of variance
Hi, I would like to estimate coefficients using poisson regression and then get standard errors that are adjusted for heteroskedasticity, using a complex sample survey data. Then I will calculate prevalence ratio and confidence intervals. Can sandwich estimator of variance be used when observations arent independent? In my case, observations are independent across groups (clusters), but not necessarily within groups. Can I calculate the standard errors with robust variance, in complex sample survey data using R? Outputs: design_tarv<-svydesign(ids=~X2, strata=~X3, data=banco, weights=~X4) banco.glm7 <- svyglm(y ~x1, data = banco, family = poisson (link= "log"), design= design_tarv) summary(banco.glm7) Call: svyglm(y ~ x1, data = banco, family = poisson(link = "log"), design = design_tarv) Survey design: svydesign(ids = ~X2, strata = ~X3, data = banco, weights = ~X4) Coefficients: Estimate Std. Error t value Pr(>|t|) (Intercept) -0.918930.04696 -19.570 < 2e-16 *** x1 0.197100.06568 3.001 0.00603 ** --- Signif. codes: 0 *** 0.001 ** 0.01 * 0.05 . 0.1 1 (Dispersion parameter for poisson family taken to be 0.5722583) Number of Fisher Scoring iterations: 5 library(sandwich) vcovHC(banco.glm7) (Intercept)x1 (Intercept) 4.806945e-13 -4.771409e-13 x1 -4.771409e-137.127168e-13 sqrt(diag(vcovHC(banco.glm7, type="HC0"))) (Intercept) x1 6.923295e-078.426314e-07 # I think this result isnt correct, because standard errors are so small. Thank you for the help, Roberta Niquini. -- ENSP - Fiocruz __ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
[R] prevalence ratio and confidence intervals
Hi everybody, I would like to estimate prevalence ratio and confidence intervals. I tried to do a log-binomial regression, but there was a failure of convergence. Now, I would like to learn how to do a poisson regression with robust variance. I am trying to estimate coefficients with poisson regression and then get standard errors that are adjusted for heteroskedasticity. glm22<- svyglm(y~x1+x2+x3+offset(log(x4)), data = banco, family = poisson, design= design_tarv) # Y has a binomial distribution (0/1) # X1, X2, X3 e X4 are categorical variables. #I am using the library(survey) because it is an analysis of Complex Sample Survey Data . summary(glm22) Call: svyglm(y~x1+x2+x3+ offset(log(x4)),data = banco, family = poisson, design = design_tarv) Survey design: svydesign(ids = ~conglomerado, strata = ~estrato, data = banco, weights = ~peso) Coefficients: Estimate Std. Error t value Pr(>|t|) (Intercept) -5.612240.07223 -77.699 < 2e-16 *** x1 0.338470.07428 4.557 0.000155 *** x2 0.177450.07059 2.514 0.019765 * x3 0.335080.09447 3.547 0.001808 ** x4 0.243820.08808 2.768 0.011217 * --- Signif. codes: 0 *** 0.001 ** 0.01 * 0.05 . 0.1 1 (Dispersion parameter for poisson family taken to be 0.7535822) Number of Fisher Scoring iterations: 5 # Using family=quasipoisson, I found the same values. library(sandwich) vcovHAC(glm22) (Intercept)x1 x2 x3x4 (Intercept)1.060857e-12-1.306035e-13-5.139155e-13 -9.788354e-13 -3.428080e-13 x1 -1.306035e-13 7.237868e-13 -3.263182e-13 -1.620593e-13 1.704392e-13 x2 -5.139155e-13 -3.263182e-13 1.250564e-12 7.207572e-13 -9.350062e-13 x3 -9.788354e-13 -1.620593e-13 7.207572e-13 1.707176e-12 -2.244859e-13 x4 -3.428080e-13 1.704392e-13 -9.350062e-13 -2.244859e-13 2.031640e-12 sqrt(diag(vcovHAC(glm22))) (Intercept) x1x2x3 x4 1.029979e-06 8.507566e-07 1.118286e-06 1.306589e-06 1.425356e-06 I think these standards errors are very small. Is this the correct form to do poisson regression with robust variance? Thank you for the help, Roberta. __ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
