data-raw/process.R

In lmiratrix/simITS: Analysis via Simulation of Interrupted Time Series (ITS) Data

# data-raw/process.R

# Process the datafiles used as illustrations in the package

library( tidyverse )




#### Make Mechlenberg data  ####

meck = read_csv( "data-raw/mech_month_outcomes.csv")
nrow( meck )
meck = rename( meck, month = month_t )


meck2 = read_csv( "data-raw/CaseLevel190322.csv" )
head( meck2 )
nrow( meck2 )
summary( meck2 )
table( meck2$Severity, useNA= "always" )



#### Make Mechlenberg subgroup data  ####

meck2.l = meck2 %>% filter( !is.na( Bail ), !is.na( Severity ) ) %>%
  group_by( month, Severity ) %>%
  summarise( n.cases = n(),
             n.bail = sum( Bail ) )

meck2.l = mutate( meck2.l, pbail = n.bail / n.cases )
head( meck2.l )

meck2.l = mutate( meck2.l, 
                  category = fct_recode( Severity,
                                         "felony" = "Felony",
                                         "misdem" = "Misdemeanor",
                                         "traffic" = "Traffic" ) )

meck2.l$Severity = NULL

if ( FALSE ) {
  meck2 = read_csv( "data-raw/PSADataforExample190212.csv")
  head( meck2 )
  
  
  pbails = grep( "pbail.", names( meck2 ) )
  ns = 3:5
  pbails
  ns
  times = gsub( "pbail.", "", names( meck2 )[pbails] )
  times
  
  meck2.l = reshape( as.data.frame( meck2 ),  idvar=c("month"),
                     varying=list( ns, pbails ),
                     timevar = "category",
                     times = times,
                     v.names=c("n","pbail"), direction="long")
  
  
  meck2.l = mutate( meck2.l, nbail = round( pbail * n.cases ),
                    pbail = pbail * n.cases / n )
  head( meck2.l )
}


if ( FALSE ) {
  meck = read_csv( "data-raw/mech_month_outcomes.csv")
  meck = rename( meck, month = month_t )
  head( meck )
  
  meck2 = read_csv( "data-raw/PSADataforExample190212.csv")
  head( meck2 )
  meck2 = mutate( meck2, tot.cases = n.felony + n.misdem + n.traffic )
  summary( meck$tot.cases - meck$n.cases )
  
  meck = merge( meck, meck2, by="month", all=TRUE )
  nrow( meck )
  meck = filter( meck, complete.cases( meck ) )
  head( meck )
  
  
  meck = mutate( meck, 
                 pbail.felony = pbail.felony * n.cases / n.felony,
                 pbail.misdem = pbail.misdem * n.cases / n.misdem,
                 pbail.traffic = pbail.traffic * n.cases / n.traffic )
  
}             

if ( FALSE ) {
  meck.l = gather( meck, pbail, pbail.felony, pbail.misdem, pbail.traffic, key="group", value="pbail" )
  
  meck.n = gather( meck, n.cases, n.felony, n.misdem, n.traffic, key="group", value="n" )
  
  ggplot( meck.l, aes( month, pbail, col=group ) ) +
    geom_line()
  
  ggplot( meck.n, aes( month, n, col=group ) ) +
    geom_line()
  
  m2 = mutate( meck, ncases2 = n.felony + n.misdem + n.traffic,
               pbail2 = (n.felony * pbail.felony + n.traffic * pbail.traffic + n.misdem * pbail.misdem) / n.cases )
  
  summary( m2$ncases - m2$ncases2 )
  
  summary( m2$pbail2 - m2$pbail )
  
}


#### Make New Jersey data #####

if ( TRUE ) {
  nj = read_csv("data-raw/njbymonth.csv" )
  nj = nj[ -c(2:11)]
  head( nj )
  nj = dplyr::select( nj, -H1, -H2 )
  summary( nj$temperature )
  sd( nj$temperature )
  
  nj = rename( nj, n.warrant = compwarr.cs,
               n.summons = compsumm.cs,
               n = comptot.cs,
               sin.m = tsin1,
               cos.m = tcos1
  )
  nj = mutate( nj, n.warrant = round( n.warrant ),
               n.summons = round( n.summons ) )
  nj$n = nj$n.warrant + nj$n.summons 
  
  with( nj, n - n.summons - n.warrant )
  #nj = dplyr::select( month, compwarr.cs, compsumm.cs, comptot.cs, temperature )
}


# Make a new (fake) new jersey -- used before data permissions were given
if ( FALSE ) {
  
  set.seed( 1019 )
  library( simITS )
  njf = make.fake.data( t.min= -7*12, t.max = 18, t0 = 0, rho = 0.4, sd.omega = 4,
                        coef.q = c( 9, 5, -12, 0 ),
                        coef.temp = 0.20,
                        coef.tx = c( 8, 0.25, -1 ) )
  head( njf )
  njf = mutate( njf, Y = round( Y * 50 ),
                Ystr = Ystr * 50,
                Ystr0 = Ystr0 * 50 )
  ggplot( data=njf, aes( month, Y ) ) +
    geom_line() +
    geom_line( aes( y = Ystr ), col="green" ) +
    geom_line( aes( y = Ystr0 ), col="blue" ) +
    geom_vline( xintercept = 0.5 )
  njf$Ystr = NULL
  
  # make the real the fake
  nj = njf
  
}


#### Save all the data to the real data for the package ####

mecklenberg = meck
newjersey = nj
meck_subgroup = meck2.l

usethis::use_data(mecklenberg, newjersey, meck_subgroup, overwrite = TRUE)


# Make skeleton documentation for the datasets
if ( FALSE ) {
  sinew::makeOxygen(mecklenberg, add_fields = "source")
  sinew::makeOxygen(newjersey, add_fields = "source")
  sinew::makeOxygen(meck_subgroup, add_fields = "source")
}