Announcements

2/8/2017 Key for Homework 3 & 4 posted

For HW3, I updated the key with the answers for problem #1.

1/26/17 Sample code to simulate AR-1 (mean-reverting)

 
R script file icon testing b estimation.R1K Download

1/26/17 Sample code to simulate AR-1 (mean-reverting)

#AR-1
#Discrete Gompertz Model
TT=1000
b=.95
u=.1
q=.2
x=rep(0,TT)
x[1]=u/(1-b) #start at equilibrium
for(t in 2:TT) x[t]=b*x[t-1]+u+rnorm(1,0,sqrt(q))

plot(x,type="l")
abline(h=u/(1-b))
title(mean(x))

#what is the mean if we de-mean the data?
mean(x-mean(x))  #it's zero

#what is u if the mean is zero? Zero

#this is why you can set u to zero in the models

1/26/17 Read up on AIC and Cross-Validation

So far we have been using AICc to compare models. It is easy and fast to compute and we wanted you to focus on model structure and the logic behind model formulation. This week, we'll introduce cross-validatio​n which is becoming more common than AICc. Read up on AIC and cross-validatio​n on Rob Hyndman's blog

1/16/17 Kalman Filter

I won't go into the Kalman Filter much in lecture. You can find many write-ups that go into great detail on it by googling.

Announcements

No announcements

Send questions about this workspace to Eli Holmes.