I am trying to find an ARIMA model for the unemployment rate in Boston after the data is differenced (code below). The goal is to have the white noise test (wbtestq) to return a probability as close to 1.000 as possible and a Q-statistic that’s as small as possible. No matter what I try I have not found a model that works. What is the correct model? clear all
freduse BOXRSA BOST625URN gen hp = BOXRSA gen unemp = BOST625URN drop if hp == . | unemp == . gen yearm = mofd(daten) gen year = year(daten) gen t = _n format yearm %tm gen month = month(daten) tab month, gen(mth) tsset yearm
qui: do http://Eduwebsite
/(3,1,3) did not work. a lot of these are converging. the middle (p,d,q) "d" should be 1 because its first differenced so that is covered, but i have not been able to find the right p and q/ arima unemp, arima (p,d,q) predict residun1, resid wntestq residun1, lags(10)
drop residun1
freduse BOXRSA BOST625URN gen hp = BOXRSA gen unemp = BOST625URN drop if hp == . | unemp == . gen yearm = mofd(daten) gen year = year(daten) gen t = _n format yearm %tm gen month = month(daten) tab month, gen(mth) tsset yearm
qui: do http://Eduwebsite
/(3,1,3) did not work. a lot of these are converging. the middle (p,d,q) "d" should be 1 because its first differenced so that is covered, but i have not been able to find the right p and q/ arima unemp, arima (p,d,q) predict residun1, resid wntestq residun1, lags(10)
drop residun1
