Quant lvl 2 - Reading 11 (Time-Series)

Question 1

Formula for simplest form of a linear trend

Answer 1

y = b0 + b1 + error1

Question 2

Define what positive exponential growth mean to the random variable for time series

Answer 2

Means that the random variable (time series) tends to increase at some constant rate of growth

Question 3

When is a log-linear model used?

Answer 3

used frequently when time series data exhibits exponential growth

Question 4

When is it appropriate to use a linear trend model?

Answer 4

when data points appear to be equally distributed above/below regression line

Question 5

does inflation rate data use linear trend models or log-linear trend usually?

Answer 5

linear trend

Question 6

When is it appropriate to use a log-linear model?

Answer 6

when the data points are curves, meaning that the residuals from the linear trend (if used) will be persistently pos. or negative for a period of time - so better to use log-linear model

Question 7

variable grows at constant rate versus constant amount, which model is appropriate for rate versus amount?

Answer 7

rate = log linear, amount = linear

Question 8

What is serial correlation?

Answer 8

limitation to trend models; means the error terms are pos. or neg. correlated for some time. The stand. errors will be unreliable and may not give correct stat. sign or insign.

Question 9

Define autoregressive models (AR)

Answer 9

“when dependent variable is regressed against one or more lagged values of itself.” (past values of variables are used to predict current/future value of variable)

Question 10

What are three conditions of covariance stationary?

Answer 10

1. “constant and finite expected value”
2. “constant and finite variance”
3. “constant and finite covariance btwn values at any given lag”

Question 11

how do you test if autocorrelations are significantly diff from zero?

Answer 11

use t-test. the Test Stat is the estimated autocorrelation / stnd error. the stnd error = 1/square-root of T (T=number of observations).

Question 12

should the autocorrelations be significantly different from 0?

Answer 12

no - and if they are not, then the AR model is correctly specified

Question 13

Is Durbin-Watson Test appropriate to test for serial correlation of error terms in AR model?

Answer 13

no, need to use t-test (autocorrelation/ (1/square root T))

Question 14

PICK UP AT 11G

Answer 14

Lesson 11G