Tables with weighted proportions (svydesign + tbl_svysummary), what am I doing wrong?

[juulia]

Hi,

It's my first post here, hopefully I'm not asking something super stupid. I'm rather new with R and having some trouble applying weights for my analysis. I'm using National Survey on Drug Use and Health 2019 dataset.

I want to create tables which include the weighted proportions for variables in wdata (risktaking, martialstatus, gender, education, agegroups, ethnicity, income, sort by MDMA use).

I tried tbl_svysummary, but obviously I'm doing something really wrong because the n in this dataset after excluding minors is around 40 000 but now in the table I'm getting from this the n goes up to hundreds of millions, makes no sense.

Is there anyone who could help me? I've been trying to figure this out by myself but ugh, I just seem to bump into more problems every time I try something.

Code:

wdata <- NSDUH_adults %>%
select(risktaking,
maritalstatus,
gender,
education,
agegroups,
ethnicity,
income,
MDMA_use,
ANALWT_C,
VESTR,
VEREP)

options(survey.lonely.psu="adjust")
wdata2 <- svydesign(ids = ~VEREP,
strata = ~VESTR,
data = wdata,
weights = ~ANALWT_C,
nest=TRUE)

wdata2 %>%
tbl_svysummary(
by = MDMA_use,
include = c(gender, risktaking, MDMA_use),
label = list(gender ~ "Gender",
risktaking ~ "Risktaking")
) %>%
add_p() %>% # comparing values by "both" column
add_overall() %>%

adding spanning header

modify_spanning_header(c("stat_1", "stat_2") ~ "MDMA")

[StatSteph]

These estimates are weighted estimates. If you look at the sum of the weight variable for your data (wdata), what is it? There are about 250 million adults in the US so I expect the weighted totals to be in that ballpark.

I'm not familiar with the gtsummary package but I use the srvyr package for survey data analysis. You might want to take a look here at a tutorial I made for that: GitHub - szimmer/tidy-survey-aapor-2021: Tidy Survey Analysis in R using the srvyr Package: AAPOR

[juulia]

Thanks! Yeah actually I realized after writing this post that I should use only percentages. So probably there's nothing wrong with the code, it's just that I should make the table so that it gives me only the percentages.

Does anybody know how to do that?

[StatSteph]

Change the statistic in tbl_summary as follows:

   tbl_svysummary(
      by = MDMA_use,
      include = c(gender, risktaking, MDMA_use),
      label = list(gender ~ "Gender",
                   risktaking ~ "Risk taking"),
      statistic=list(all_categorical()~"{p}%" ) ########
   )

[juulia]

Thank you so much! That solved it!