It Just Takes a Question:
The Development of
tidyDisasters
and
tidyYRBS

Catalina Cañizares. MSc; Mark J. Macgowan. Ph.D.; Gabriel Odom. Ph.D., Th.D.

Agenda

  • Background and question
    • The data
  • The tidyDisasters package
    • The wrangling process for disasters
    • The merge
    • Examples using tidyDisasters
  • The tidyYRBS package
    • Overview of the wrangling process
    • Example using tidyYRBS
  • Future work

Background

Growing evidence suggests that surviving a natural disaster or a terrorist attack event increases the likelihood of presenting a mental health disorders, disruptive psychological functioning and may increase the likelihood of suicidaliy (Fitzpatrick & Spialek, 2020; Tang et al., 2018; Zakour, 2019).

Is there any association between mass casualty events and adolescent suicide ideation and suicide attempts?

The Data

Disaster Datasets

  • Disaster Declaration Summaries
    • Federal Emergency Management Agency (FEMA) (United States of America)
  • Emergency Events Database (EMDAT)
    • Centre for Research on the Epidemiology of Disasters (Belgium)
  • Global Terrorism Database (GTD)
    • National Consortium for the Study of Terrorism and Responses to Terrorism (United States of America)

Adolescent Suicidality Dataset

  • Youth Risk Behavior Surveillance System (YRBSS) High School Combined Dataset
    • Centers for Disease Control and Prevention
    • (Q26) During the past 12 months, did you ever seriously consider attempting suicide?
    • (Q28) During the past 12 months, how many times did you actually attempt suicide?

tidyDisasters

Disaster Datasets

FEMA

ID:
2ece821d-39e1-40a1-9556-31ab914686fa
Variables of interest:
State
County
Start date
End date
Disaster type
Challenge:
No way to filter by the number of people killed
ID with no information

EMDAT

ID:
1990-0707-USA
Variables of interest:
State
County
Start date
End date
Disaster type
Number of people killed
Number of people wounded
Challenge:
States and counties smashed in a single cell and missing locations and end dates.

GTD

ID:
201606120001
Variables of interest:
State
County
Start date
End date
Disaster type
Number of people killed
Number of people wounded

Taking a look at each data set

FEMA 60,499 X 6

# A tibble: 7 × 8
  id       disas…¹ state place…² disas…³ incidentBeginDate   incidentEndDate    
  <chr>      <dbl> <chr>   <dbl> <chr>   <dttm>              <dttm>             
1 2ece821…     852 WA      99067 DR      1990-01-06 00:00:00 1990-01-14 00:00:00
2 7ccbd3b…     852 WA      99005 DR      1990-01-06 00:00:00 1990-01-14 00:00:00
3 2768b59…     852 WA      99069 DR      1990-01-06 00:00:00 1990-01-14 00:00:00
4 fd19e41…     852 WA      99053 DR      1990-01-06 00:00:00 1990-01-14 00:00:00
5 45c152c…     852 WA      99033 DR      1990-01-06 00:00:00 1990-01-14 00:00:00
6 0b84703…     852 WA      99027 DR      1990-01-06 00:00:00 1990-01-14 00:00:00
7 8341af9…     852 WA      99041 DR      1990-01-06 00:00:00 1990-01-14 00:00:00
# … with 1 more variable: incidentType <chr>, and abbreviated variable names
#   ¹​disasterNumber, ²​placeCode, ³​disasterType

EMDAT 2,766 X 12

# A tibble: 1 × 11
  `Dis No`      Disast…¹ Locat…² Start…³ Start…⁴ Start…⁵ End Y…⁶ End M…⁷ End D…⁸
  <chr>         <chr>    <chr>     <dbl>   <dbl>   <dbl>   <dbl>   <dbl>   <dbl>
1 1990-0707-USA Flood    Washin…    1990       1       9    1990       1      11
# … with 2 more variables: `Total Deaths` <dbl>, `No Injured` <dbl>, and
#   abbreviated variable names ¹​`Disaster Type`, ²​Location, ³​`Start Year`,
#   ⁴​`Start Month`, ⁵​`Start Day`, ⁶​`End Year`, ⁷​`End Month`, ⁸​`End Day`

The Location Vector

[1] "Washington, Oregon, Idaho, Montana, Wyoming, Colorado"

The GTD 39 X 12

# A tibble: 1 × 8
       eventid iyear imonth  iday provstate city    nkill nwound
         <dbl> <dbl>  <dbl> <dbl> <chr>     <chr>   <dbl>  <dbl>
1 201606120001  2016      6    12 Florida   Orlando    50     53

The merge process was based on data equivalence criteria

  • State
  • County
  • Date or range of dates

Cleaning the locations vector

  • Separated the content from the Location vector
  • Removed unnecessary information such as “Province”
  • Mapped the states to state abbreviations
# Example of separation and unnecessary information
emdatWrangled1_df %>%
  select(event_id, Location) %>%
  mutate(LocationOrig = Location) %>%
  separate_rows(Location, sep = ",|;") %>%
  mutate(Location = str_to_lower(Location)) %>%
  mutate(
    Location = str_replace(
      Location,
      pattern = "provinces", replacement = "province"
    )
  ) %>%
  mutate(
    Location = str_remove(Location, pattern = "province")
  ) %>%
  mutate(Location = str_trim(Location, side = "both")) %>%
  mutate(
    Location = str_remove(Location, pattern = " \\)")
    
# Mapped the states to state abbreviations
emdatLocWrangled1_df <- 
    emdatWrangled1_df %>% 
    select(event_id, Location) %>% 
    mutate(LocationOrig = Location) %>% 
    mutate(
     # replacePattern1_char is a dictionary created previously
        Location = str_replace_all(Location, replacePattern1_char) 
    ) %>% 
    mutate(Location = str_to_lower(Location))

Cleaning the locations vector

  • Fixed spelling mistakes
emdatLocWrangled2_df <- 
    emdatLocWrangled1_df %>% 
    
    ###  Spelling issues  ###
    mutate(
        Location = str_replace_all(
            Location, pattern = "georgie", replacement = "georgia"
        ),
        Location = str_replace_all(
            Location, pattern = "virginie", replacement = "virginia"
        ),
        Location = str_replace_all(
            Location, pattern = "virgina", replacement = "virginia"
        ),
        Location = str_replace_all(
            Location, pattern = "caroline du nord", replacement = "north carolina"
        ), 
        Location = str_replace_all(
            Location, pattern = "north craolina", replacement = "north carolina"
        ),
        Location = str_replace_all(
            Location, pattern = "nth.carolina", replacement = "north carolina"
        ),
        Location = str_replace_all(
            Location, pattern = "norh carolina", replacement = "north carolina"
        ),
        Location = str_replace_all(
            Location, pattern = "californie", replacement = "california"
        ),
        Location = str_replace_all(
            Location, pattern = "louisiane", replacement = "louisiana"
        ),
        Location = str_replace_all(
            Location, pattern = "lousiana", replacement = "louisiana"
        ),
        # MS smart quotes strike again
        Location = str_replace_all(
            Location, pattern = "hawai‘i", replacement = "hawaii"
        ),
        Location = str_replace_all(
            Location, pattern = "hawai", replacement = "hawaii"
        ),
        Location = str_replace_all(
            Location, pattern = "hawaiii", replacement = "hawaii"
        ),
        Location = str_replace_all(
            Location, pattern = "okhlahoma", replacement = "oklahoma"
        ),
        Location = str_replace_all(
            Location, pattern = "ohlahoma", replacement = "oklahoma"
        ),
        Location = str_replace_all(
          Location, pattern = "okhahoma", replacement = "oklahoma"
        ),
        Location = str_replace_all(
            Location, pattern = "washigton", replacement = "washington"
        ),
        Location = str_replace_all(
            Location, pattern = "wsahington", replacement = "washington"
        ),
        Location = str_replace_all(
            Location, pattern = "mississipi", replacement = "mississippi"
        ),
        Location = str_replace_all(
            Location, pattern = "mississppi", replacement = "mississippi"
        ),
        Location = str_replace_all(
            Location, pattern = "mssissippi", replacement = "mississippi"
        ),
        Location = str_replace_all(
            Location, pattern = "pennsylvannia", replacement = "pennsylvania"
        ),
        Location = str_replace_all(
            Location, pattern = "pensylvania", replacement = "pennsylvania"
        ),
        Location = str_replace_all(
            Location, pattern = "pennsylvanie", replacement = "pennsylvania"
        ),
        Location = str_replace_all(
            Location, pattern = "pennyslvania", replacement = "pennsylvania"
        ),
        Location = str_replace_all(
            Location, pattern = "nebarska", replacement = "nebraska"
        ),
        Location = str_replace_all(
            Location, pattern = "ilinois", replacement = "illinois"
        ),
        Location = str_replace_all(
            Location, pattern = "new mexique", replacement = "new mexico"
        ),
        Location = str_replace_all(
            Location, pattern = "missourri", replacement = "missouri"
        ),
        Location = str_replace_all(
            Location, pattern = "arkanasas", replacement = "arkansas"
        ),
        Location = str_replace_all(
            Location, pattern = "arkensas", replacement = "arkansas"
        ),
        Location = str_replace_all(
            Location, pattern = "tannesse", replacement = "tennessee"
        ),
        Location = str_replace_all(
            Location, pattern = "tennesse", replacement = "tennessee"
        ),
        Location = str_replace_all(
            Location, pattern = "tennesseee", replacement = "tennessee"
        ),
        Location = str_replace_all(
            Location, pattern = "kensas", replacement = "kansas"
        ),
        Location = str_replace_all(
            Location, pattern = "hansas", replacement = "kansas"
        ),
        Location = str_replace_all(
            Location, pattern = "massachusets", replacement = "massachusetts"
        ),
        Location = str_replace_all(
            Location, pattern = "massachussetts", replacement = "massachusetts"
        ),
        Location = str_replace_all(
          Location, pattern = "massachussets", replacement = "massachusetts"
        ),
        Location = str_replace_all(
            Location, pattern = "minissotta", replacement = "minnesota"
        ),
        Location = str_replace_all(
            Location, pattern = "minnessota", replacement = "minnesota"
        ),
        Location = str_replace_all(
            Location, pattern = "indianaa", replacement = "indiana"
        ),
        Location = str_replace_all(
            Location, pattern = "indiania", replacement = "indiana"
        ),
        Location = str_replace_all(
            Location, pattern = "montania", replacement = "montana"
        ),
        Location = str_replace_all(
            Location, pattern = "rhode islands", replacement = "rhode island"
        ),
        Location = str_replace_all(
            Location, pattern = "sth.carolina", replacement = "SC"
        ),
        Location = str_replace_all(
            Location, pattern = "floride", replacement = "florida"
        ),
        Location = str_replace_all(
            Location, pattern = "new jesey", replacement = "new jersey"
        ),
        Location = str_replace_all(
            Location, pattern = "south dacota", replacement = "south dakota"
        )
    )

Cleaning the locations vector

  • Added states where only the city was mentioned
  • Eliminated directional words
  • Replace New England with the actual states
  • Separated the vector regions into two columns, one that contained only the state abbreviations and the other one the rest of the information (city or county)
# Added states where only the city was mentioned 
 mutate(
    Location = case_when(
        event_id == "1991-0426-USA" ~ "oakland (CA)",
      TRUE ~ Location
    ),
    Location = case_when(
        event_id == "1994-0149-USA" ~
          "los angeles (CA)",
        TRUE ~ Location
      ),
    Location = case_when(
      event_id == "1995-0288-USA" ~
          "atlanta (GA), detroit (MI), philadelphia (PA), chicago (IL), milwaukee (WI)",
        TRUE ~ Location
    ),
    Location = case_when(
      event_id == "1995-0073-USA" ~
        "los angeles (CA), dallas (TX), fort worth (TX), NM",
      TRUE ~ Location
    )
    
###  Remove Directional Words  ###
# Eliminate words "southern", "south", "eastern", "north", "west", "western"
#   "area", "near", "areas", "northern"
emdatLocWrangled2_df <- 
    emdatLocWrangled_df %>% 
    mutate(
        Location = str_remove_all(Location, pattern = "central")
    ) %>% 
    mutate(
        Location = str_remove_all(Location, pattern = "southern")
    ) %>% 
    mutate(
        Location = str_remove_all(Location, pattern = "south")
    ) %>% 
    mutate(
        Location = str_remove_all(Location, pattern = "eastern")
    ) %>% 
    mutate(
        Location = str_remove_all(Location, pattern = "east")
    

# Replaced New England with the actual states
mutate (
    Location = str_replace_all(
     Location, pattern = "new england", 
      replacement = "RI, CT, MA, NH, VT, ME"
    )
  ) 


# Split state abbreviations from regions
    rowwise() %>% 
    mutate(
        State = case_when(
            !inProgress ~ Location,
            inProgress ~ paste(
                unlist(
                    str_extract_all(
                        Location, pattern = "[A-Z]{2}"
                    )
                ),
                collapse = ", "
            )
        )
    ) %>% 
    mutate(
        Region = case_when(
            !inProgress ~ NA_character_,
            inProgress ~ paste(
                unlist(
                    str_remove_all(
                        Location, pattern = "[A-Z]{2}"
                    )
                ),
                collapse = ", "
            )
        )
    ) %>% 
    ungroup()

The resulting data set

From here

# A tibble: 762 × 2
   `Dis No`      Location                                                       
   <chr>         <chr>                                                          
 1 1990-0346-USA Oklahoma, Kansas, Nebraska, Missouri, Iowa, Illinois, Alabama,…
 2 1990-0357-USA Texas, Oklahoma, Arkansas, Virginie, Caroline du Nord          
 3 1990-0359-USA Texas, Oklahoma, Arkansas, Kansas, Missouri, Illinois, Indiana…
 4 1990-0360-USA Nebraska, Iowa, Kansas, Missouri, Texas, Oklahoma, Arkansas    
 5 1990-0363-USA Colorado, Kansas, Missouri, Indiana, Kentucky                  
 6 1990-0350-USA Texas                                                          
 7 1990-0353-USA Oklahoma                                                       
 8 1990-0355-USA Texas, Kansas, Louisiane, Alabama, Georgie, Caroline du Sud    
 9 1990-0362-USA Indiana, Illinois, Ohio, Kansas, Kentucky                      
10 1990-0707-USA Washington, Oregon, Idaho, Montana, Wyoming, Colorado          
# … with 752 more rows

To here

# A tibble: 1,369 × 3
   event_id      State County        
   <chr>         <chr> <chr>         
 1 1990-0073-USA OH    belmont       
 2 1991-0192-USA CA    los angeles   
 3 1991-0218-USA NY    richmond      
 4 1991-0426-USA CA    alameda       
 5 1992-0120-USA CA    humboldt      
 6 1992-0215-USA CA    san bernardino
 7 1992-0271-USA HI    kauai         
 8 1992-0271-USA HI    honolulu      
 9 1993-0126-USA NE    boyd          
10 1993-0126-USA NE    butler        
# … with 1,359 more rows

Dates

  • We pivot longer the dates to guarantee more matches when adding EMDAT and FEMA together
  • Corrected dates where the end happened before the beginning
###### Pivot the dates to make the ranges long ################################

emdatclean_final<- emdatLocClean3_df %>% 
  mutate(diff=  end_date- begin_date) %>% 
  group_by(`region_id`) %>%
  mutate(Day = list(seq (begin_date, end_date, by = 'day'))) %>% 
  unnest(Day) %>%
  ungroup() %>% 
  select(Day, diff, begin_date, end_date, everything())

# We have some events where the end happens before the beginning
femaRaw1_df %>% 
    mutate(duration = end_date - begin_date) %>% 
    # rowwise() %>% 
    mutate(duration = end_date - begin_date) %>% 
    filter(duration < 0)

femaRaw2_df <-
    femaRaw1_df %>% 
    mutate(
    # Hurricane Earl: https://www.fema.gov/disaster/1241
    begin_date = case_when(
      eventid == "DR-1241-FL" ~ as_date("1998-09-03"),
      TRUE ~ begin_date
    ),
    end_date = case_when(
      eventid == "DR-1241-FL" ~ as_date("1998-09-04"),
      TRUE ~ end_date
      
 ###  Imputation by Mean  ###
      
ImputeMean <- function(x) {
    replace(x, is.na(x), mean(x, na.rm = TRUE))
}
# A tibble: 16,167 × 4
   Day        diff    begin_date end_date  
   <date>     <drtn>  <date>     <date>    
 1 1990-06-21  0 days 1990-06-21 1990-06-21
 2 1990-07-01 27 days 1990-07-01 1990-07-28
 3 1990-07-02 27 days 1990-07-01 1990-07-28
 4 1990-07-03 27 days 1990-07-01 1990-07-28
 5 1990-07-04 27 days 1990-07-01 1990-07-28
 6 1990-07-05 27 days 1990-07-01 1990-07-28
 7 1990-07-06 27 days 1990-07-01 1990-07-28
 8 1990-07-07 27 days 1990-07-01 1990-07-28
 9 1990-07-08 27 days 1990-07-01 1990-07-28
10 1990-07-09 27 days 1990-07-01 1990-07-28
# … with 16,157 more rows

Merge

allEventsMap_df <-
    full_join(
        fema_df %>%
            select(Day, state, county, femaID = event_id),
        emdat_df %>%
            select(Day, state, county, emdatID = event_id),
        by = c("Day", "state", "county")
    )

36,224,806 x 5

# A tibble: 14 × 5
   Day        state county     femaID           emdatID           
   <date>     <chr> <chr>      <chr>            <chr>             
 1 2008-04-04 TX    san saba   EM-3284-TX_43119 2008-0131-USA_0011
 2 2008-04-08 TX    crosby     EM-3284-TX_11121 2008-0131-USA_0011
 3 1998-06-07 FL    manatee    DR-1223-FL_1011  1998-0185-USA_0001
 4 2008-04-15 TX    brazos     EM-3284-TX_2424  2008-0131-USA_0011
 5 2005-12-28 TX    san saba   DR-1624-TX_47129 2005-0724-USA_0001
 6 2008-04-04 TX    guadalupe  EM-3284-TX_19500 2008-0131-USA_0011
 7 2006-01-01 TX    harris     DR-1624-TX_22625 2005-0724-USA_0001
 8 2005-12-29 TX    swisher    DR-1624-TX_50469 2005-0724-USA_0001
 9 2008-04-03 TX    jeff davis EM-3284-TX_25413 2008-0131-USA_0011
10 2011-06-05 TX    mason      DR-1999-TX_9514  2011-0639-USA_0003
11 2008-04-02 TX    ochiltree  EM-3284-TX_40318 2008-0131-USA_0011
12 2008-03-19 TX    jones      EM-3284-TX_26062 2008-0131-USA_0011
13 2011-04-24 TX    wheeler    DR-1999-TX_13219 2011-0140-USA_0001
14 2008-03-21 TX    williamson EM-3284-TX_50365 2008-0131-USA_0011

The key

This key contains the beginning year, month, state, and a ticker to differentiate events with the same year, month and state.

# A tibble: 7 × 4
  femaID        emdatID            smashedID                    eventKey     
  <chr>         <chr>              <chr>                        <chr>        
1 DR-852-WA_001 1990-0707-USA_0001 DR-852-WA_1990-0707-USA_0001 1990_01_WA_02
2 DR-852-WA_002 1990-0707-USA_0001 DR-852-WA_1990-0707-USA_0001 1990_01_WA_02
3 DR-852-WA_003 1990-0707-USA_0001 DR-852-WA_1990-0707-USA_0001 1990_01_WA_02
4 DR-852-WA_004 1990-0707-USA_0001 DR-852-WA_1990-0707-USA_0001 1990_01_WA_02
5 DR-852-WA_005 1990-0707-USA_0001 DR-852-WA_1990-0707-USA_0001 1990_01_WA_02
6 DR-852-WA_006 1990-0707-USA_0001 DR-852-WA_1990-0707-USA_0001 1990_01_WA_02
7 DR-852-WA_007 1990-0707-USA_0001 DR-852-WA_1990-0707-USA_0001 1990_01_WA_02

Another challenge

FEMA and EMDAT had different ways of categorizing and naming the types of natural disaster.
Therefore:

# A tibble: 10,412 × 4
   eventKey      incident_type    hazard_cluster        hazard_type             
   <chr>         <chr>            <chr>                 <chr>                   
 1 1990_01_MS_01 Severe Storms    Precipitation-related METEOROLOGICAL and HYDR…
 2 1990_01_OR_02 Flood            Flood                 METEOROLOGICAL and HYDR…
 3 1990_01_WA_02 Flood            Flood                 METEOROLOGICAL and HYDR…
 4 1990_02_AL_01 Severe Storms    Precipitation-related METEOROLOGICAL and HYDR…
 5 1990_02_GA_01 Severe Storms    Precipitation-related METEOROLOGICAL and HYDR…
 6 1990_02_IL_01 Severe Ice Storm Precipitation-related METEOROLOGICAL and HYDR…
 7 1990_02_TN_01 Flood            Flood                 METEOROLOGICAL and HYDR…
 8 1990_03_AL_02 Severe Storms    Precipitation-related METEOROLOGICAL and HYDR…
 9 1990_03_FL_01 Flood            Flood                 METEOROLOGICAL and HYDR…
10 1990_03_NY_01 Snow             Precipitation-related METEOROLOGICAL and HYDR…
# … with 10,402 more rows

Disaster types

Fifth normal form

  • Remove Redundancy
  • Memory Efficiency

Data in the tidyDisasters Package

-allKeys_df
-disastCasualties_df
-disastDates_df
-disastLocations_df
-disastTypes_df
-emdat_hazard_cluster_df
-fema_hazard_cluster_df
-hazard_report_df

Example # 1 using the tidyDisasters package

# Can we find the largest Ice Storm since the 90's?

library(tidyDisasters)
data("disastLocations_df")
data("disastTypes_df")
data("disastDates_df")

disastLocations_df %>%
  left_join(disastTypes_df) %>%
  left_join(disastDates_df) %>%
  mutate(Year = year(eventStart)) %>% 
  filter(incident_type %in% c("Severe Ice Storm", "Snow")) %>%
  group_by(state, county, Year) %>%
  summarise(Frozen = n() >= 1L, .groups = "keep") %>%
  group_by(Year) %>%
  summarise(Count = sum(Frozen)) %>%
  ggplot() +
  aes(x = Year, y = Count) +
  labs(
    title = "Number of Counties Affected by Snow Storms Since the 90s",
    caption = "Data from the tidyDisasters R Package",
    y = "No. Counties Frozen"
  ) +
  scale_x_continuous(breaks = seq(1990,2020,5)) +
  theme_classic() +
  geom_point() +
  gghighlight(Count > 400)

We were able to plot and find the 1993 “Storm of the Century” that affected almost 800 counties and the 1996 blizzard that affected around 500 counties and is remembered as the greatest snowstorm in terms of the amount of snow that fell.

Example # 2 using the tidyDisasters package

# Are disasters becoming more prevalent?

data("disastTypes_df")
data("disastDates_df") 

disastTypes_df %>%
  left_join(disastDates_df) %>%
  mutate(Year = year(eventStart)) %>% 
  filter(
    hazard_cluster %in% c(
      "Precipitation-related", " Wind-related",
      "Environmental degradation (Forestry)", "Flood", "Convective-related",
      "Pressure-related"
    )
  ) %>%
  group_by(Year, hazard_cluster) %>%
  count(hazard_cluster) %>%
  ggplot() +
  aes(x = Year, y = n, color = hazard_cluster) +
  labs(
    title = "Disasters tendency since 1990",
    caption = "Data from the tidyDisasters R Package",
    y = "Count"
  ) +
  geom_point() +
  facet_wrap(~hazard_cluster, scales = "free_y") +
  theme_classic() +
  theme(legend.position = "none")

This plot evidences the trend on six different types of hazard clusters.

tidyYRBS

The process to get the data into R and Wrangle

  • Modified line 318 from the spss syntax because it exceed the number of characters allowed and it generated an error
  • Cleaned the sitename vector
  • Renamed sex variables and corrected the type of vector
  • We ignored the variable “qntransgender” because it s almost entirely missing
  • Cleaned the names of variables

Created subsets of interest

Suicidality

# A tibble: 235,761 × 6
   record is_hopeless suicide_considered suicide_planned suicide_attem…¹ suici…²
   <chr>  <lgl>       <lgl>              <lgl>                     <int> <lgl>  
 1 120    TRUE        FALSE              TRUE                          0 FALSE  
 2 121    FALSE       FALSE              FALSE                         0 FALSE  
 3 123    FALSE       FALSE              FALSE                         0 FALSE  
 4 125    TRUE        TRUE               TRUE                          3 TRUE   
 5 126    TRUE        TRUE               TRUE                          6 TRUE   
 6 127    TRUE        TRUE               TRUE                          1 FALSE  
 7 5864   TRUE        TRUE               FALSE                         0 FALSE  
 8 7370   FALSE       TRUE               TRUE                          1 FALSE  
 9 7371   FALSE       FALSE              FALSE                         0 FALSE  
10 9659   TRUE        FALSE              FALSE                         0 FALSE  
# … with 235,751 more rows, and abbreviated variable names ¹​suicide_attempts,
#   ²​suicide_injury

Substance use

# A tibble: 33,634 × 10
   record nAlcoholDays nMariju…¹ used_…² used_…³ used_…⁴ used_…⁵ used_…⁶ used_…⁷
   <chr>         <dbl>     <dbl> <lgl>   <lgl>   <lgl>   <lgl>   <lgl>   <lgl>  
 1 136             0        0    FALSE   FALSE   FALSE   FALSE   FALSE   FALSE  
 2 137             0        0    FALSE   FALSE   FALSE   FALSE   FALSE   FALSE  
 3 3976            1.5      4.5  FALSE   FALSE   FALSE   FALSE   FALSE   FALSE  
 4 5027            1.5      1.12 TRUE    TRUE    TRUE    TRUE    TRUE    TRUE   
 5 5871            1.5      1.12 TRUE    TRUE    FALSE   TRUE    TRUE    TRUE   
 6 7382            0        0    FALSE   FALSE   FALSE   FALSE   FALSE   FALSE  
 7 7383            0        0    FALSE   FALSE   FALSE   FALSE   FALSE   FALSE  
 8 9664            1.5      1.12 FALSE   FALSE   TRUE    FALSE   FALSE   FALSE  
 9 9667            0        1.12 FALSE   FALSE   FALSE   FALSE   FALSE   FALSE  
10 9668            7.5      4.5  TRUE    TRUE    FALSE   TRUE    TRUE    TRUE   
# … with 33,624 more rows, 1 more variable: used_injection <lgl>, and
#   abbreviated variable names ¹​nMarijuanaDays, ²​used_prescrip_pain,
#   ³​used_cocaine, ⁴​used_inhalant, ⁵​used_heroin, ⁶​used_methamphetamine,
#   ⁷​used_MDMA

Data in the tidyYRBS Package

  • hs_district
  • hs_demographics
  • hs_suicide
  • hs_substance_use
  • clean_yrbs_2019
  • raw_yrbs_2019

Example using the tidyYRBS

data(hs_substance_use)

the_data <- left_join(hs_demographics, hs_substance_use)

# Weights
the_data_weights <- the_data |>
  srvyr::as_survey_design(
    ids=PSU,
    weights=weight,
    strata=stratum,
    nest = TRUE
  )

alc_by_sex <- the_data_weights |>
  mutate(
    used_alcohol = case_when(
      nAlcoholDays == 0 ~ 0,
      nAlcoholDays > 0 ~ 1
    )
   ) |>
  filter(Grade == 12) |>
  group_by(year, Sex) |>
  summarise(prevalence = mean(used_alcohol, na.rm=T),
            n = n()
  ) |>
  filter(!is.na(Sex))

alc_by_sex <- the_data |>
  mutate(
    used_alcohol = case_when(
      nAlcoholDays == 0 ~ 0,
      nAlcoholDays > 0 ~ 1)) |>
  filter(Grade == 12) |>
  group_by(year, Sex) |>
  summarise(prevalence = mean(used_alcohol, na.rm=T)) |>
  filter(!is.na(Sex))
   
ggplot(alc_by_sex, aes(year, prevalence, color=Sex, label=Sex)) +
  geom_smooth(alpha=0.1, size=0) +
  geom_textline() +
  scale_color_manual(values=c("#420039", "indianred1")) +
  theme_minimal(base_size = 12) +
  theme(legend.position = "none") +
  theme(panel.grid.major = element_blank(), panel.grid.minor = element_blank()) +
  scale_x_continuous(breaks = seq(1990,2020,2)) +
  theme(axis.text.x = element_text(angle = 90)) +
  scale_y_continuous(lim=c(.25,.6),
                     breaks = seq(0,1,0.05),
                     labels = scales::percent) +
  labs(y="Alcohol use prevalence", x="",
       title="Any alcohol use",
       caption = "Grade 12th. YRBS, 1990-2019")

Future work

  • Respond our original question
  • Submit tidyDisasters to CRAN
  • Create the relational datasets for each health behavior measured by the YRBS
    • Behaviors that contribute to unintentional injuries and violence
    • Sexual behaviors related to unintended pregnancy and sexually transmitted infections
    • Tobacco use
    • Unhealthy dietary behaviors
    • Inadequate physical activity
  • Submit tidyYRBS to CRAN

The End

2 years in 20 minutes…

tidyDisasters and tidyYRBS