The Basics: Outline

What is a time series? What is a financial time series? What is the purpose of our analysis? Classification of Time Series. Correlation
Autocorrelation Partial Autocorrelation Cross Correlation

Basic transformation to stationarity

Differencing
Spring 2005 K. Ensor, STAT 421

What is a time series?

Review
Random variable Distribution (cdf, pdf) Moments
Mean Variance Covariance Correlation Skewness Kurtosis

Time Series
Random process random variable is a function of time Distribution? Moments

K. Ensor, STAT 421

Mean Variance Covariance Correlation Skewness Kurtosis

Spring 2005

Stationary Time Series with 100 Future Realizations

-4
0

-2

10

20

30

Spring 2005

K. Ensor, STAT 421

U.S. Weekly Interest Rates Red line is 1-y ear; Blue line is 3-y ear
16 percent 4 6 8 10 12 14

01/05/1962

07/18/1969

01/28/1977

08/10/1984

02/21/1992

09/03/1999

Spring 2005

Time in weeks K. Ensor, STAT 421

Further examples of a time series

Anything observed sequentially (by time?) Returns, volatility, interest rates, exchange rates, bond yields, Hourly temperature, hourly ozone levels ???
Spring 2005
Quarterly Earning per shar Johnson and Johnson

earning 0 5

10

15

Jan 60

Jan 64

Jan 68

Jan 72

Jan 76

Jan 80

Time in quarters

K. Ensor, STAT 421

What is different?
The observations are not independent. There is correlation from observation to observation. Consider the log of the J&J series. Is there correlation in the observations over time?
Spring 2005

Log Quarterly Earning per share Johnson and Johnson

log earning

0
Jan 60

Jan 64

Jan 68

Jan 72

Jan 76

Jan 80

Time in quarters

K. Ensor, STAT 421

What are our objectives?

Making decisions based on the observed realization requires:
Descriptive: Estimating summary measures (e.g. mean) Inferential: Understanding / Modeling Prediction / Forecasting Control of the process

If correlation is present between the observations then our typical approaches are not correct (as they assume iid samples).
Spring 2005 K. Ensor, STAT 421

Classification of a Time Series

Dimension of T
Time, space, spacetime

Dimension of X
Univariate Multivariate

Nature of T
Discrete
Equally Unequally spaced

State spce
Discrete Continuous

Memory types
Stationary
No memory Short memory Long memory

Continuous
Observed continuously Observed by some random process
Spring 2005

Nonstationary

K. Ensor, STAT 421

Stationarity

Strictly Stationary
All finite dimensional distributions are the same.

Covariance Stationary
First and second moment structure does not change with time.

Spring 2005

K. Ensor, STAT 421

Autocorrelation

Spring 2005

K. Ensor, STAT 421

10

Autocorrelation Function for a CSTS

In theory

How to estimate this quantity?

Spring 2005 K. Ensor, STAT 421

11

Autocorrelation?

Lagged Scatterplots : x

How would you determine or show correlation over time?

Series 1

Series 1

Series 1

Series 1

Series 1
0 1 lagged 4 2

1 lagged 1

1 lagged 2

1 lagged 3

1 lagged 5

Series 1

Series 1

Series 1

Series 1

Series 1
0 1 lagged 9 2

1 lagged 6

1 lagged 7

1 lagged 8

1 lagged 10

Spring 2005

K. Ensor, STAT 421

12

Sample ACF and PACF

Sample ACF sample estimate of the autocorrelation function.
Substitute sample estimates of the covariance between X(t) and X(t+h). Note: We do not have n pairs but n-h pairs. Subsitute sample estimate of variance.

Sample PACF correlation between observations X(t) and X(t+h) after removing the linear relationship of all observations in that fall between X(t) and X(t+h).
Spring 2005 K. Ensor, STAT 421

13

Summary Plots
Log Quarterly Earnings f or J&J
2 0 1

Jan 60

Jan 64

Jan 68 Time in quarters

Jan 72

Jan 76

Jan 80

Histogram
1.0 15

ACF
1.0

PACF

0.5

10

ACF

ACF
0 5 10 Lag 15

0.0

-0.5

-1.0

-1

-1.0
0

-0.5

0.0

0.5

10 Lag

15

Log Quarterly Earnings f or J&J

Spring 2005

K. Ensor, STAT 421

14

Cross Correlation

Spring 2005

K. Ensor, STAT 421

15

Multivariate Series
How can we study the relationship between 2 or more time series? U.S. weekly interest rate series measured in percentages
Time: From 1/5/1962 to 9/10/1999. Variables: r1(t) = The 1-year Treasury constant maturity rate r2(t) = The 3-year Treasury constant maturity rate
Spring 2005 K. Ensor, STAT 421

And the corresponding change series

c1(t)=(1-B)r1(t) c2(t)=(1-B)r2(t)

16

U.S. Weekly Interest Rates Red line is 1-year; Blue line is 3-year
16 percent 4
01/05/1962

10

12

14

07/18/1969

01/28/1977

08/10/1984

02/21/1992

09/03/1999

Time in weeks

Spring 2005

K. Ensor, STAT 421

17

Scatterplot of U.S. Weekly Interest Rate

Scatterplot of Change 1yr rate and 3yr rate

1 .5 d iff( yr 3 )

Scatterplots between series simultaneous in time and the change in each series. The two series are highly correlated.

16

12

yr 3

10

10 y r1

12

14

16

- 1 .0

- 0 .5

0 .0

0 .5

1 .0

14

- 1 .5

- 1 .0

- 0 .5

0 .0 d iff( y r 1 )

0 .5

1 .0

1 .5

Spring 2005

K. Ensor, STAT 421

18

Change in 1-year rate

1.5 -1.5 -0.5 0.5

500

1000 Time

1500

2000

Change in 3-year rate

1.5 -1.0 0.0 0.5 1.0

500

1000

1500

2000

Spring 2005

K. Ensor, STAT 421

19

Scatterplot of Change in 1-year and 3-year rate

1.5 c3 -1.0 0.0 0.5 1.0

-1.5

-1.0

-0.5

0.0 c1

0.5

1.0

1.5

Multivariate Series : cbind(c1, c3)

0.8 1.0

c1
0.8

c1 and c3

What is the cross-correlation between the two series?

ACF 0.2 0.4 0.6

0.0

10

15

20

25

30

0.0
0

0.2

0.4

0.6

10

15

20

25

30

0.8

ACF 0.4 0.6

0.2

0.0

-30

-25

-20

-15 Lag

-10

-5

0.0

0.2

0.4

0.6

0.8

1.0

c3 and c1

c3

10

15 Lag

20

25

30

Spring 2005

K. Ensor, STAT 421

20

Differencing to achieve Stationarity

Spring 2005

K. Ensor, STAT 421

21

Detrending by taking first difference.

Y(t)=X(t) X(t-1)
What happens to the trend? Suppose X(t)=a+bt+Z(t) Z(t) is a random variable.
0.2 0.4

First Difference Log Quarterly Earning per share J&J

first difference of log earning

-0.6
Apr 60

-0.4

-0.2

0.0

Apr 64

Apr 68

Apr 72

Apr 76

Apr 80

Time in quarters

Spring 2005

K. Ensor, STAT 421

22

Sumary Plots of Detrended J&J log earnings per share.

Detrended Log Quarterly Earnings for J&J
0.4 -0.6
Apr 60

0.0

Apr 64

Apr 68

Apr 72

Apr 76

Apr 80

Time in quarters
Histogram
1.0

ACF
1.0

PACF

30

0.5

20

ACF

ACF
0 5 10 Lag 15

0.0

10

-0.5

-1.0

-0.8

-0.6

-0.4

-0.2

0.0

0.2

0.4

0.6

-1.0
0

-0.5

0.0

0.5

10 Lag

15

Detrended Log Quarterly Earnings f or J&J

Spring 2005

K. Ensor, STAT 421

23

Removing Seasonal Trend one way to proceed.

Suppose Y(t)=g(t)+W(t) where g(t)=g(t-s) where s is our season for all t. W(t) is again a new random variable Form a new series U(t) by taking the s difference
U(t)=Y(t)-Y(t-s) =g(t)-g(t-s) + W(t)-W(t-s) =W(t)-W(t-s) again a random variable
Spring 2005 K. Ensor, STAT 421

24

Summary of Transformed J&J Series

Log J&J Af ter Remov ing Linear and Seasonal Trend
0.2 -0.2 -0.1 0.0 0.1

Apr 61

Jan 65

Oct 68 Time in quarters

Jul 72

Apr 76

Jan 80

Histogram
1.0 15

ACF
1.0

PACF

0.5

10

ACF

ACF
0 5 10 Lag 15

0.0

-0.5

-1.0

-0.2

-0.1

0.0

0.1

0.2

0.3

-1.0
0

-0.5

0.0

0.5

5 Lag

10

15

Log J&J Af ter Remov ing Linear and Seasonal Trend

Spring 2005

K. Ensor, STAT 421

25

Summary of Transformations:
X(t) = log (Q(t)) Y(t)=X(t)-X(t-1) = (1-B)X(t) U(t)= (1-B4)Y(t) U(t)=(1-B4) (1-B)X(t)

Spring 2005

K. Ensor, STAT 421

26

An example of Forecasting

Spring 2005

K. Ensor, STAT 421

27

What is the next step?

U(t) is a time series process called a moving average of order 1 (or possibly a MA(1) plus a seasonal MA(1))
U(t)=q e(t-1) + e(t)

Proceed to estimate q and then we can estimate summary information about the earnings per share as well as predict the future earnings per share.

Spring 2005

K. Ensor, STAT 421

28

Forecast of J&J series

Two Year Forecast and 95% Bounds for Johnson and Johnson Quarterly Earnings Per Share

Earnings

10

15

20

25

10 Quarter Time

15

20

Spring 2005

K. Ensor, STAT 421

29

Wrap up
Basics of distribution theory. Classification of time series. Basics of stationarity. Correlation functions
Autocorrelation Partial autocorrelation Cross correlation

Transformations to a stationary series

differencing
Spring 2005 K. Ensor, STAT 421

30