CODEWITHSUNDEEP
rlogistic-regressionmissing-dataglmr-mice
Wael Hussein
Wael Hussein

Reputation: 155

Odd ratios and 95% confidence intervals from logistic regression on data imputed with MICE

I have data with missing values which I imputed using the MICE package. 

impData <- mice(analysis_set,m=5,maxit=50,meth='pmm',seed=500)

Now I need to run logistic regression analysis:

modelFit1 <- with(data = impData, 
exp = glm(formula = Outcome ~ inputVar1 + inputVar2 + inputVar3, 
family = binomial(link = "logit")))

I can get a pooled analysis using:

pool(modelFit1)

And more info using:

summary(pool(modelFit1))

The last command shows the estimates, SE, t, df, Pr(>|t|), lo 95, hi 95, nmis, fmi and lambda.

My question is: is there an easy way to obtain the ORs and 95% CI from the pooled analysis?

I used to do that on a dataset using:

exp(cbind(OR = coef(mylogit), confint(mylogit)))

where mylogit is the glm() for a dataset. Is there an equivalent to that for the pooled analysis?

Upvotes: 1

Views: 4128

Answers (2)

PP Bao
PP Bao

Reputation: 21

The pooled object is under the class mipo from MICE package.

summary(pool(modelFit1), conf.int = TRUE, exponentiate = TRUE)

gives both Odds Ratio (estimate) and corresponding CIs.

Upvotes: 2

Wael Hussein
Wael Hussein

Reputation: 155

Thanks @user20650

summaryPool <- summary(pool(modelFit1))
exp(cbind(summaryPoolM2[,1],summaryPoolM2[,6],summaryPoolM2[,7]))

Column 1 is the estimates, 6 and 7 are the ln(95% confidence intervals). Exponentiating these values gives the ORs and 95% CIs.

Upvotes: 1

Related Questions