Off by one hotfix

R/retrieve.R

@@ -1,11 +1,10 @@
 #' @title Get hospitalization data
-#' @name Get flu and COVID-19 hospitalization data from healthdata.gov endpoint
 #' @description Retrieves hospitalization data from the healthdata.gov endpoint with optional filtering on fields, and return the results into a nice tibble.
 #' @param endpoint URL to healthdata.gov endpoint (see references)..
 #' @param state A two-letter state abbreviation.
 #' @param limitrows Limit API query to at most this number of results. Default is `NULL` (no limit).
-#' @param mindate Minimum date of results returned (ISO 8601 format: YYYY-MM-DD). See examples.
-#' @param maxdate Maximum date of results returned (ISO 8601 format: YYYY-MM-DD). See examples.
+#' @param mindate Minimum date of results returned (ISO 8601 format: YYYY-MM-DD). See examples. Note that after retrieving the data, dates will be shifted back one day because the admissions data is _previous_ day's flu/covid admissions.
+#' @param maxdate Maximum date of results returned (ISO 8601 format: YYYY-MM-DD). See examples. Note that after retrieving the data, dates will be shifted back one day because the admissions data is _previous_ day's flu/covid admissions.
 #' @param limitcols Limit the columns returned to the subjectively defined important ones?
 #' @param app_token App token from healthdata.gov. If `NULL` you might get rate limited. Add an entry to your `~/.Renviron` with `HEALTHDATA_APP_TOKEN="tokenhere"` that you got from <https://healthdata.gov/profile/edit/developer_settings>.
 #' @return A tibble
@@ -89,6 +88,9 @@ get_hdgov_hosp <- function(endpoint="https://healthdata.gov/resource/g62h-syeh.j
     d <- d %>% dplyr::select(state:cov.deaths.cov)
   }
 
+  # Data is previous day's stats. Shift dates back one day to deal with this.
+  d$date <- d$date-1
+
   message(paste0(nrow(d), " rows retrieved from:\n", api_url))
   return(d)
 

scratch/eval.R

@@ -32,7 +32,7 @@ eval_wrap <- function(hosp = NULL, ilidat = NULL, min_hhs = "2020-10-12", max_hh
 
     hosp <-
       get_hdgov_hosp(mindate = min_hhs, maxdate=max_hhs) %>%
-      mutate(date = date - 1) %>%
+      # mutate(date = date - 1) %>%
       mutate(flu.admits = as.numeric(flu.admits),
              flu.admits.cov = as.numeric(flu.admits.cov)) %>%
       mutate(epiweek = lubridate::epiweek(date),
@@ -128,7 +128,7 @@ models <-
 ## alternative is to keep hosp=NULL param to eval_wrap and the function will retrieve each time
 hosp_dat <-
   get_hdgov_hosp(mindate =  "2020-10-12", maxdate="2021-12-27") %>%
-  mutate(date = date - 1) %>%
+  # mutate(date = date - 1) %>%
   mutate(flu.admits = as.numeric(flu.admits),
          flu.admits.cov = as.numeric(flu.admits.cov)) %>%
   mutate(epiweek = lubridate::epiweek(date),

scratch/explore-covariates.R

@@ -26,7 +26,7 @@ tail(h)
 # weekly data summed across states
 hu <-
   h %>%
-  mutate(date = date - 1) %>%
+  # mutate(date = date - 1) %>%
   mutate(epiweek = lubridate::epiweek(date),
          epiyear = lubridate::epiyear(date),
          .after=date) %>%

scratch/hosp-arima.R

@@ -3,7 +3,7 @@ library(fiphde)
 
 tmp_hosp_dat <-
   get_hdgov_hosp(maxdate="2021-12-26") %>%
-  mutate(date = date - 1) %>%
+  # mutate(date = date - 1) %>%
   mutate(flu.admits = as.numeric(flu.admits),
          flu.admits.cov = as.numeric(flu.admits.cov)) %>%
   mutate(epiweek = lubridate::epiweek(date),

scratch/workflow.R

@@ -37,7 +37,7 @@ hosp <- get_hdgov_hosp(maxdate="2021-12-26")
 
 tmp_weekly_flu <-
   hosp %>%
-  mutate(date = date - 1) %>%
+  # mutate(date = date - 1) %>%
   mutate(flu.admits = as.numeric(flu.admits),
          flu.admits.cov = as.numeric(flu.admits.cov)) %>%
   mutate(epiweek = lubridate::epiweek(date),