HHS region ili forecasting

[stephenturner]

Forecasting ILI was brought up in #7, solved in #31, (but without florida, see #27). However, some of the state-level forecasts are poor.

Explore modeling HHS regions instead of state-level regions.

1. This will require modifying get_cdc_ili. Currently this takes either state or national. cdcfluview::ilinet() will allow for "hhs" region, but the joins that get_cdc_ili() does downstream will break if using hhs regions.
2. If forecasting seems more reliable at a national level, it will require some bookkeeping (in another issue) to associate states with their respective HHS region, and apply the correct hhs regional model to each state.

[stephenturner]

Working in #46. Reliability for hhs region 2 is suspect.

[stephenturner]

Thanks @vpnagraj for sending me to the cdc surveillance repo. Found something in there that'll be useful, https://github.com/CDCgov/SARS-CoV-2_Genomic_Surveillance/blob/78c48739ca32d9f44667e0c3702ee2ec09e78562/Code/variant%20surveillance%20system.R#L21-L31

HHS_reg = list(HHS1=c("CT", "ME", "MA", "NH", "RI", "VT"),
               HHS2=c("NJ", "NY", "PR", "VI"),
               HHS3=c("DE", "DC", "MD", "PA", "VA", "WV"),
               HHS4=c("AL", "FL", "GA", "KY", "MS", "NC", "SC", "TN"),
               HHS5=c("IL", "IN", "MI", "MN", "OH", "WI"),
               HHS6=c("AR", "LA", "NM", "OK", "TX"),
               HHS7=c("IA", "KS", "MO", "NE"),
               HHS8=c("CO", "MT", "ND", "SD", "UT", "WY"),
               HHS9=c("AZ", "CA", "HI", "NV", "AS", "MP", "GU", "MH"),
               HHS10=c("AK", "ID", "OR", "WA")
)

Should we ever want to tackle what's mentioned ☝️

2. If forecasting seems more reliable at a national level, it will require some bookkeeping (in another issue) to associate states with their respective HHS region, and apply the correct hhs regional model to each state.