Working with margins of error in tidycensus

Analyzing US Census Data in R

Kyle Walker

Instructor

ACS data vs. census data

Decennial census provides official counts
ACS provides population characteristics
Learn more from the ACS handbook

Margins of error in the ACS

get_acs(geography = "county", 
        variables = c(median_age = "B01002_001"), 
        state = "OR")

# A tibble: 36 x 5
   GEOID NAME                     variable   estimate   moe
   <chr> <chr>                    <chr>         <dbl> <dbl>
 1 41001 Baker County, Oregon     median_age     48.2   0.4
 2 41003 Benton County, Oregon    median_age     32.6   0.3
 3 41005 Clackamas County, Oregon median_age     41.4   0.2
 4 41007 Clatsop County, Oregon   median_age     43.7   0.4
 5 41009 Columbia County, Oregon  median_age     43.3   0.4
 6 41011 Coos County, Oregon      median_age     48.2   0.3
 7 41013 Crook County, Oregon     median_age     48.3   0.7
# ... with 29 more rows

Inspecting margins of error

vt_eldpov <- get_acs(geography = "tract", 
                     variables = c(eldpovm = "B17001_016", 
                                   eldpovf = "B17001_030"), 
                     state = "VT")
vt_eldpov

# A tibble: 368 x 5
   GEOID       NAME                 variable estimate   moe
   <chr>       <chr>                <chr>       <dbl> <dbl>
 1 50001960100 Census Tract 9601... eldpovm        0.    9.
 2 50001960100 Census Tract 9601... eldpovf        5.    5.
 3 50001960200 Census Tract 9602... eldpovm        0.    9.
 4 50001960200 Census Tract 9602... eldpovf        0.    9.
 5 50001960300 Census Tract 9603... eldpovm       16.   14.
 6 50001960300 Census Tract 9603... eldpovf        5.    7.
# ... with 362 more rows

Using margin of error functions

tidycensus includes these functions for calculating margins of error:

moe_sum(): MOE for a derived sum
moe_product(): MOE for a derived product
moe_ratio(): MOE for a derived ratio
moe_prop(): MOE for a derived proportion

Group-wise margins of error

vt_eldpov2 <- vt_eldpov %>%
  group_by(GEOID) %>%
  summarize(
    estmf = sum(estimate), 
    moemf = moe_sum(moe = moe, estimate = estimate))
vt_eldpov2

# A tibble: 184 x 3
   GEOID       estmf moemf
   <chr>       <dbl> <dbl>
 1 50001960100     5 10.3 
 2 50001960200     0  9   
 3 50001960300    21 15.7 
 4 50001960400    29 11.4 
 5 50001960500     0  9   
# ... with 179 more rows

Let's practice!

Analyzing US Census Data in R