export.R

# Export

ggsave("output/tno_trust.png", plot = tno_trust)
ggsave("output/rheum_trust.png", plot = rheum_trust)
ggsave("output/pain_trust.png", plot = pain_trust)
ggsave("output/all_MSK_trust.png", plot = all_MSK_trust)
ggsave("output/NCL_spec.png", plot = NCL_MSK_trust)
ggsave("output/NCL_MSK_all.png", plot= NCL_MSK_all_trust)

NCL_MSK_trust
all_MSK_trust
NCL_MSK_all_trust

write_csv(tno_hw, file = "output/tno_forecast.csv")
write_csv(rheum_hw, file = "output/rheum_forecast.csv")
write_csv(pain_hw, file = "output/pain_forecast.csv")
write_csv(all_MSK_hw, file = "output/all_MSK_forecast.csv")
write_csv(NCL_MSK_hw, file = "output/NCL_MSK_forecast.csv")
write_csv(NCL_MSK_all_hw, file = "output/NCL_MSK_all_forecast.csv")


library(openxlsx)

openxlsx::write.xlsx(list("tno" = tno_hw
                          , "rheum" = rheum_hw
                          , "pain" = pain_hw
                          , "All" = all_MSK_hw
                          , "NCL" = NCL_MSK_hw
                          , "Total_NCL" = NCL_MSK_all_hw)
                     , "output/MSK_OP_forecast.xlsx")



#trend change
library(broom)

tno_mods <-
  tno_hw %>% 
  group_by(Trust) %>% 
  select(Trust, .mean, Date) %>% 
  do(model = lm(.mean ~ Date, data = .))


# pull out coefficients
tno_subs <- lapply(tno_mods[[2]], coef)

rheum_mods <-
  rheum_hw %>% 
  group_by(Trust) %>% 
  select(Trust, .mean, Date) %>% 
  do(model = lm(.mean ~ Date, data = .))

rheum_subs <- lapply(rheum_mods[[2]], coef)

pain_mods <-
  pain_hw %>% 
  group_by(Trust) %>% 
  select(Trust, .mean, Date) %>% 
  do(model = lm(.mean ~ Date, data = .))

pain_subs <- lapply(pain_mods[[2]], coef)

all_MSK_mods <-
  all_MSK_hw %>% 
  group_by(Trust) %>% 
  select(Trust, .mean, Date) %>% 
  do(model = lm(.mean ~ Date, data = .))

all_MSK_subs <- lapply(all_MSK_mods[[2]], coef)

NCL_mods <-
  NCL_MSK_hw %>% 
  group_by(Specialty) %>% 
  select(.mean, Date) %>% 
  do(model = lm(.mean ~ Date, data = .))

NCL_MSK_subs <- lapply(NCL_mods[[2]], coef)

NCL_mods_all <-  lm(.mean ~ Date, data = NCL_MSK_all_hw)


# Assemble into table
per_growth <-
  data.frame(
    Trust = c(tno_mods[[1]])
    , tno = sapply(tno_subs, "[[", 2)  
    , rheum = sapply(rheum_subs, "[[", 2)  
    , pain = sapply(pain_subs, "[[", 2)  
    , all = sapply(all_MSK_subs, "[[", 2)
  )

per_growth <-
  rbind(per_growth,
        data.frame(
          Trust = "All",
          tno = NCL_MSK_subs[[3]][[2]],
          rheum = NCL_MSK_subs[[2]][[2]],
          pain = NCL_MSK_subs[[1]][[2]],
          all = coef(NCL_mods_all)[2]
        ))

per_growth 




# Replacement section for the error over the coefficients above.
# Rationale is change from month1 to month 60 in prediction
# Percentage change: (month60 - month1) / month1

# Total
NCL_MSK_all_hw %>% 
  slice_head() %>% 
  select(.mean) 
#2431

NCL_MSK_all_hw %>% 
  slice_tail() %>% 
  select(.mean) 
# 2656

(2656 - 2431) / 2431

# Function for pulling out fist and last and doing percentage change.
fun <-
  function(x, max_index){
    (x[[max_index]]-x[[1]])/x[[1]]
  }

# Tno
tno_preds <-  lapply(tno_mods[[2]], predict)

tno_perc_change <- lapply(tno_preds, lm_perc_change_fun, max_index = 60)

# pain
pain_preds <-  lapply(pain_mods[[2]], predict)

pain_perc_change <- lapply(pain_preds, lm_perc_change_fun, max_index = 60)

# Rheum
rheum_preds <-  lapply(rheum_mods[[2]], predict)

rheum_perc_change <- lapply(rheum_preds, lm_perc_change_fun, max_index = 60)

# all_trust
all_MSK_mods_preds <-  lapply(all_MSK_mods[[2]], predict)

all_MSK_mods_perc_change <- lapply(all_MSK_mods_preds, lm_perc_change_fun, max_index = 60)

# NCL
NCL_preds <-  lapply(NCL_mods[[2]], predict)

NCL_perc_change <- lapply(NCL_preds, lm_perc_change_fun, max_index = 60)


# NCL total
NCL_all_preds <-  predict(NCL_mods_all)

NCL_all_perc_change <- lm_perc_change_fun(NCL_all_preds, max_index = 60)





unlist(tno_perc_change)


# Assemble into table
per_growth_2<-
  data.frame(
    Trust = c(tno_mods[[1]])
    , tno = unlist(tno_perc_change)  
    , rheum = unlist(rheum_perc_change)  
    , pain = unlist(pain_perc_change) 
    , all = unlist(all_MSK_mods_perc_change)
  )

per_growth_2 <-
  rbind(per_growth_2,
        data.frame(
          Trust = "All",
          tno = NCL_perc_change[[1]],
          rheum = NCL_perc_change[[2]],
          pain = NCL_perc_change[[3]],
          all = NCL_all_perc_change 
        ))

per_growth_2