alex/Population_Study
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Working for the day, switch to ARMA

Browse Source
This commit is contained in:
Alex Gebben Work 2025-09-30 17:08:53 -06:00
parent f46ed27939
commit 5122fe26b6
2 changed files with 57 additions and 7 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

22
ARMA_Pop.r Normal file
View File

@ -0,0 +1,22 @@
library(tidyverse)
library(forecast)
source("Scripts/Functions.r")
#source("Scripts/Load_Wyoming_Web_Data.r")
DF <- FRED_GET('WYLINC3POP','LN_POP') %>% select(-YEAR)
TS <- 1000*ts(DF,start=c(1970),end=c(2024),frequency=1)
BC <- BoxCox.lambda(TS)
MODEL <- auto.arima(TS, lambda = BC)
plot(forecast(MODEL,h=35),main="Lincoln County Population Forecast")
#####Plan and ideas
#1) Review IMPLAN for industry multipliers
#2) Review IMPLAN for employment to population multipliers (imparted)
#3) Find a list of all planned new projects
#4) Use the IMPLAN multipliers for each sector to estimate total change
#5) Develop survey to estimate likelihood of new projects
#6) Compare to the ARMA percentile
#7) Adjust the ARMA up assuming some of these outputs are known.
####Other ideas, develop larger plan? Maybe look at decline in other industries as a proportion of employment

42
BVAR_Pop.r
View File

@ -33,6 +33,7 @@ DATA_TO_GATHER[[length(DATA_TO_GATHER)+1]] <- c('IDCARI9POP','CARIBOU_POP',FALS
DATA_TO_GATHER[[length(DATA_TO_GATHER)+1]] <- c('IDBONN0POP','BONNEVILLE_POP',FALSE)
###US Population
DATA_TO_GATHER[[length(DATA_TO_GATHER)+1]] <- c('POPTOTUSA647NWDB','US_POP',FALSE)
DATA_TO_GATHER[[length(DATA_TO_GATHER)+1]] <- c('CE16OV','US_EMP',FALSE)
@ -44,21 +45,48 @@ for(x in 1:length(DATA_TO_GATHER)){
if(x==1){RES <- C_DATA}else{RES <- RES %>% full_join(C_DATA)}
rm(CURRENT,C_DATA)
}
DATA <- RES %>% mutate(US_POP=US_POP-WY_POP,WY_POP=WY_POP-LN_POP-UINTA_POP-SUBLETTE_POP-SWEETWATER_POP-TETON_POP)
colnames(DATA)
TS_DATA_ORIG <- DATA %>% select(YEAR,LN_POP,US_POP,WY_POP,UINTA_POP,SUBLETTE_POP,SWEETWATER_POP,TETON_POP,BEAR_LAKE_POP,CARIBOU_POP,BONNEVILLE_POP) %>% filter(!is.na(LN_POP)) %>% arrange(YEAR) %>% select(-YEAR)
TS_DATA_ORIG <- DATA %>% select(YEAR,LN_POP,LN_EMPLOYMENT,US_EMP,US_POP,WY_POP,UINTA_POP,SUBLETTE_POP,SWEETWATER_POP,TETON_POP,BEAR_LAKE_POP,CARIBOU_POP,BONNEVILLE_POP) %>%
filter(!is.na(LN_POP),!is.na(LN_EMPLOYMENT)) %>%
arrange(YEAR) %>% select(-YEAR)
TEST <- RES %>% filter(!is.na(LN_EMPLOYMENT)) %>% mutate(LAG_LN_EMP=lag(LN_EMPLOYMENT))
TEST %>% select(LN_EMPLOYMENT,LAG_LN_EMP)
feols(log(LN_POP)~log(LAG_LN_EMP)+lag(LN_POP)+YEAR,data=TEST )
colnames(DATA)
TS_DATA <- log(ts(TS_DATA_ORIG,start=c(1970),end=c(2024),frequency=1))
plot(TS_DATA)
?bv_minnesota
MOD <- bvar(TS_DATA,lags=2, n_draw=15000)
tsplot(TS_DATA)
library(fixest)
TEMP <- feols(log(LN_POP) ~ lag(LN_POP)+log(US_EMP)+lag(log(US_EMP))+log(US_POP),data=DATA)
TEM
plot(TEMP$residuals)
MOD <- bvar(TS_DATA,exogen="sdfs",sdflkj=5,lags=2, n_draw=15000)
plot(predict(MOD,horizon=25,value="LN_POP"),area=TRUE)
forecast(MOD,variables=c("LN_POP"),horizon=25)
?predict.bvar
?bv_mh
summary(MOD)
plot(MOD)
plot(fitted(MOD,type="mean"))
plot(residuals(MOD,type="mean"),vars=c("LN_POP","UINTA_POP","SWEETWATER_POP"))
plot(MOD, type = "dens", vars_response = "LN_POP", vars_impulse = "LN_POP-lag1")
opt_irf <- bv_irf(horizon = 25, identification = TRUE)
plot(irf(MOD,opt_irf,conf_bands = c(0.05, 0.1,0.15)),area=TRUE,vars_response = c("LN_POP"),vars_impulse = c("UINTA_POP","SWEETWATER_POP","WY_POP"))
plot(predict(MOD,horizon=25,conf_bands = c(0.05, 0.1,0.15)),area=TRUE,vars=c("LN_POP","UINTA_POP"))
plot(irf(MOD,opt_irf,conf_bands = c(0.05, 0.1,0.15)),area=TRUE,vars_impulse = c("LN_EMP"),vars_response = c("WY_POP","LN_POP","UINTA_POP","SUBLETTE_POP","SWEETWATER_POP"))
plot(irf(MOD,opt_irf,conf_bands = c(0.05, 0.1,0.15)),area=TRUE,vars_impulse = c("LN_EMP"),vars_response = c("TETON_POP","BEAR_LAKE_POP","CARIBOU_POP","BONNEVILLE_POP"))
plot(irf(MOD,opt_irf,conf_bands = c(0.05, 0.1,0.15)),area=TRUE,vars_impulse = c("LN_POP"),vars_response = c("WY_POP","LN_POP","UINTA_POP","SUBLETTE_POP","SWEETWATER_POP"))
plot(irf(MOD,opt_irf,conf_bands = c(0.05, 0.1,0.15)),area=TRUE,vars_impulse = c("LN_POP"),vars_response = c("TETON_POP","BEAR_LAKE_POP","CARIBOU_POP","BONNEVILLE_POP"))
DATA2 <- RES %>% mutate(US_POP=US_POP-WY_POP,WY_POP=WY_POP-LN_POP-UINTA_POP-SUBLETTE_POP-SWEETWATER_POP-TETON_POP)
TS_DATA2 <- DATA2 %>% select(YEAR,LN_POP,US_POP,WY_POP,UINTA_POP,SUBLETTE_POP,SWEETWATER_POP,TETON_POP,BEAR_LAKE_POP,CARIBOU_POP,BONNEVILLE_POP) %>%
dplyr::filter(!is.na(LN_POP)) %>%
arrange(YEAR) %>% select(-YEAR) %>% ts %>% log
MOD2 <- bvar(TS_DATA2,lags=5, n_draw=15000)
plot(irf(MOD2,opt_irf,conf_bands = c(0.05, 0.1,0.15)),area=TRUE,vars_response = c("LN_POP"))
?plot.bvar_irf
plot(predict(MOD,horizon=25,conf_bands = c(0.05, 0.1,0.15)),area=TRUE,vars=c("LN_POP"))
exp(3.5)-exp(3)
acf(resid(MOD))