I have performed an unadjusted logistic regression using weights (obtained via genetic matching) as below. I am using the survey package to make working with the weights easier:
fit <- svyglm(outcome ~ group, design = match_df_svy, family = quasibinomial(link = "logit"))
ShowRegTable(fit)
exp(coef) [confint] p
(Intercept) 0.11 [0.09, 0.14] <0.001
groupTRUE 1.14 [0.88, 1.46] 0.323
group represents my intervention. Am I interpreting this correctly that since this is an unadjusted regression, the intercept in this case would represent the odds ratio when group = FALSE?
This doesn't seem to be the case by just looking at the incidence of the outcome in both groups:
group outcomeFALSE outcomeTRUE se.outcomeFALSE se.outcomeTRUE
FALSE FALSE 0.8980632 0.1019368 0.009810129 0.009810129
TRUE TRUE 0.8858308 0.1141692 0.007180766 0.007180766
A rough calculation of the odds ratio from this 2x2 table results in similar point estimates and CIs:
epitools::oddsratio(c(898, 101, 885, 114))
$data
Outcome
Predictor Disease1 Disease2 Total
Exposed1 898 101 999
Exposed2 885 114 999
Total 1783 215 1998
$measure
odds ratio with 95% C.I.
Predictor estimate lower upper
Exposed1 1.000000 NA NA
Exposed2 1.144994 0.8623838 1.522286
$p.value
two-sided
Predictor midp.exact fisher.exact chi.square
Exposed1 NA NA NA
Exposed2 0.3492472 0.3863462 0.3479743
$correction
[1] FALSE
attr(,"method")
[1] "median-unbiased estimate & mid-p exact CI"
I assume therefore that I am interpreting what the intercept actually means incorrectly and am just looking for some basic guidance.
