SE.estimate and CI

[lourens]

Hi All,
I might have a very simple question, but it is giving me some trouble. I have run secr models incorporating sex as a covariate in both sigma and p, as CL=True

Code: Select all

Msex2.lap<-secr.fit(lap.ch, model = list(g0~Sex, sigma~Sex), mask=mask.lap,CL=T, trace = FALSE)

I then use to

Code: Select all

derived(Msex2.lap, alpha = 0.05, se.D = TRUE))

to get density and SE estimates. My question is related to SE and the lcl and ucl estimated

estimate SE.estimate lcl ucl CVn CVa CVD
D 5.175057e -04 0.0002825136 0.0001901883 0.001408142 0.3336084 0.4321198 0.545914

From my understanding 95 % CI can also be computed as , mean +-1.96*SE. My qeustion is why does the CI estimated here differ from the 'manual' calculation? Or am I completely off the ball here? If I am correct, which value should I use for my CI?
Thanks for response.
Lourens

[murray.efford]

Hi Lourens
Asymmetric confidence intervals (symmetric +/- 1.96 SE on the log scale) are computed by default. Asymmetric intervals are generally thought to have better properties (i.e. better coverage). You can choose a symmetric interval if you wish:

Code: Select all

derived(secrdemo.0)
     estimate SE.estimate      lcl      ucl       CVn        CVa       CVD
esa 13.869099          NA       NA       NA        NA         NA        NA
D    5.479808   0.6445538 4.354997 6.895136 0.1147079 0.02602642 0.1176234

derived(secrdemo.0, loginterval = F)
     estimate SE.estimate      lcl     ucl       CVn        CVa       CVD
esa 13.869099          NA       NA      NA        NA         NA        NA
D    5.479808   0.6445538 4.216506 6.74311 0.1147079 0.02602642 0.1176234

You may also want to think about whether to specify distribution = 'binomial'; this is justified if you want an interval for the 'realized' density which is the same as assuming fixed N in the masked area.
Murray

[lourens]

Hi Murray,
Thanks so much; yes if run the symmetric interval it gives me similar results to my manual calcuation. It is so weird, I just went through the manual again and saw the loginteval script. As they say read the manual....
Thanks again.
Lourens