[R-sig-Epi] R Survey Package - Confidence Intervals and Odds ratios

[peter dalgaard]

Not my area of expertise really, but offhand: If summary.svyglm wants a df argument, so probably does confint.svyglm. 

Quick check says that it has one, except that it is called ddf=...

It takes quite a low df (2 or so) to get a discrepancy that large, though, so maybe you should share the output of summary() and degf()?

- pd

> On 3 Mar 2022, at 03:22 , Ashley Splain <asplain1 using umbc.edu> wrote:
> 
> I hope that this email finds everyone well.  I am a new R user and I am
> writing to ask if anyone may be able to provide some statistical expertise:
> 
> I am running logistic regression models using the Survey Package in R (my
> dataset requires the incorporation of survey weights). The calculation for
> the odds ratios and confidence intervals is from the basic stats package.
> However, the confidence intervals don't seem to match up with the
> significance levels - the confidence intervals suggest something is
> significant but the p level is not significant.  For example, b = -0.81, p
> = .094, OR = 0.45, CI = 0.26-0.75.
> 
> Here is the syntax for the design of the overall analysis and an example of
> one model:
> nhanesDesigni<-
> svydesign(id=~PSU,Strata=~STRA,weights=~WTMEC6YR,nest=TRUE,survey.lonely.psu="adjust",
> data=nhanesAnalysis)
> 
> actimeR1_i <-
> (svyglm(riskaccuracy~AcTime4+Age+Gender+educationR+Race5R+Insuranced+smokingdR+alcdR+Waistc,family="binomial",
> design=nhanesDesigni))
> summary(actimeR1_i,df=degf(nhanesDesigni))
> exp(coef(actimeR1_i))
> exp(confint(actimeR1_i))
> 
> Would anyone by chance know where I might be going wrong and how to rectify
> the code to obtain accurate CIs? I would appreciate any input you can
> provide!
> 
> Best,
> Ashley
> 
> 	[[alternative HTML version deleted]]
> 
> _______________________________________________
> R-sig-Epi using r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-sig-epi

-- 
Peter Dalgaard, Professor,
Center for Statistics, Copenhagen Business School
Solbjerg Plads 3, 2000 Frederiksberg, Denmark
Phone: (+45)38153501
Office: A 4.23
Email: pd.mes using cbs.dk  Priv: PDalgd using gmail.com