More on regression

a time series must be

covariance stationary

define covariance stationary

mean and the covariance is constant over time

if a time series has same mean and variance without seasonality

we can assume it is covariance stationary

is autoregression model a type of time series model?

yes, therefore it is covariance stationary

An accurately specified autoregression model will have residual autocorrelation close

to 0

if today’s stock is strongly correlated with yesterday’s stock, this is an example of

autocorrelation

the original data has autocorrelation, does the autoregression model has auto correlation?

no the auto regression model removes all autocorrelation. therefore the residual of a correct model has 0 auto correlation

what test to calculate the autocorrelation of residual

do t test

if t test < critical value

non of the autocorrelation of residual is critically different then 0, then your model is good to go

is a random walk covariance stationary

NO; so we can not do regression

all random walk have a unit root because the B1 =

1; B0= 0

what measure identifies outliers

standardized residual

and observation is influential if the standardized residual

is greater than certain critical value of t statistic with (n-k-2) degrees of freedom

how to calculate log liner model

plug in the numbers as usual, press second +ln to get the final anwser