Time Series Analysis

Arrangement of statistical data in chronological order in accordance with occurrence of time is known as Time series.

 The objective of time series analysis is to find a pattern in the historical data and then extrapolate the pattern into the future.

Time Series Patterns

i.e Actual = Pattern + error

Utilities: Determine the type and nature of the variations. Compare the actual value and the expected value. Predict the behavior of the variable in the future. Compare the changes in the values of different phenomenon at different times.

Components of a Time Series

 Trend we mean average tendency to increase or decrease over a long period of time. The oscillatory movements with period more than one year called cyclical fluctuations. Cyclical movements do not follow any regular pattern but move in a somewhat unpredictable pattern. Seasonal variation involves patterns of changes with in a year that operate in a regular and periodic manner. Irregular variations are unpredictable and are beyond the control of human hand.

Decomposition Models
Additive Model
In this model, it is assumed that the effect of various components can be estimated by adding the various components of a time-series. It is stated as: Y=T+C+S+I Where Y is the time series value at time t T is the trend value C is the cyclical variation at time t S is the seasonal variation at time t I is random fluctuation at time t
.

 Multiplicative Model
The actual values of a time series, represented by Y can be found by multiplying four components at a particular time period. The effect of four components on the time series is interdependent. The multiplicative time series model is defined as: Y=TCSI

Smoothing Methods
The objective of smoothing methods is to smooth out the random variations due to irregular components of the time series.
1. 2.

Moving averages Weighted moving averages

Moving Averages
Moving average, MA t + 1 = S{Dt + Dt - 1 + Dt n
2

+ ... + Dt -

n + 1}

where t = current time period D = actual data which is exchanged each period n = length of time period

 The following table shows the production volume in 000 tonnes for a product. Use these data to compute a 3 year moving average for all available years. And also find the error. Year Production
1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 21 22 23 25 24 22 25 26 27 26

Weighted Moving Averages

Weighted Moving average !

?( weight for period n) * (Data value in period n)A

 Vacuum cleaner sales for 12 months is given below. The owner of the supermarket decides to forecast sales by weighting the past three months as follows:
Weight applied 3 2 1
Month 1

Month Last month Two months ago Three months ago

3 13

4 16

5 19

6 23

7 26

8 30

9 28

10 18

11 16

12 14

Sales

10 12

 A state government is studying the number of traffic fatalities in the state resulting from drunken driving for each of the last 12 months.

1 2 3 4 5 6 7 8 9 10 11 12 Accidents 280 300 280 280 270 240 230 230 220 200 210 200
Month
Calculate 4 monthly moving average.

 Trend Analysis:
 It allows us to describe a historical pattern.  It permits us to project past patterns into the future.  It allows us to eliminate the trend component from the series.

 Measurement of Trend 1. Principle of least squares Linear trend Quadratic trend Exponential trend 2. Method of Semi- averages

If number of years is odd, then x t  middle year ut ! timeinterval If number of years is even, then x t  average of two middle years ut ! time interval 2

Linear ternd : y t ! b 0  b1 x t y t ! b 0  b 1u t where x t  middle year ut ! , timeinterv al x t  average of two middle years ut ! time interval 2

The normal equations to determine the unknown constants are :

! nb 0  b1 u t y t ! nb 0

y t v u t !b 0 u t  b1 u t2 y t v u t !b1 u t2

Parabolic Trend : y t ! b 0  b 1 x t  b 2 x t2 y t ! b 0  b 1u t  b 2u t2 The normal equations to determine unknown constants are :

y ! nb  b u  b u y ! nb  b u
t 0 1 t 2 t t 0 2

2 t

u u

v y t ! b 0 u t  b 1 u t2  b 2 u 3 u t v y t ! b 1 u t2 t t
2 t

v y t ! b 0 u t2  b 1 u t3  b 2 u 4 u t2 v y t ! b 0 u t2 b 2 u 4 t t

E po e tial tre

:y ! e 1 t 0 t

y ! e 1 t 0 t aski g ' log' o ot si es log(y t ) ! log( 0 )  Yt ! A  B W ere A ! log( 0 ) B! ! A tilog(A) B loge 1 ! ; loge ! 0.4343 1 loge
0 1 t

loge
t

log(y t ) ! log( 0 )   1 loge

 Plot the time series and comment on the appropriateness of a linear trend for the following data. If not what type of functional form do you believe would be most appropriate for the trend pattern of this time series?

Year Sales

10

400 390 320 340 270 260 300 320 340 370

 The president of a small manufacturing firm is concerned about the continual increase in manufacturing costs over the past several years. The following figures provide a time series of the cost per unit for the firm s leading product over the eight years.
Year Cost/unit
1 20 2 24.5 3 28.2 4 27.5 5 26.6 6 30 7 31 8 36

Does a linear trend appear to be present?

 Below are given the figures of production in 000 quintals of a sugar factory.

Year Production

1995 1996 1997 1998 1999 2000 2001 80 90 92 83 94 99 92

Fit a straight line trend Plot these data and show the trend line Estimate the production in 2004

Semi-Averages
In this method, the whole data is divided into two parts with respect to time. In case of odd number of years the two parts are obtained by omitting the value corresponding to the middle year. We compute the A.M for each part and plot these two averages against the middle values of the respective periods by each part. The line obtained on joining these two points is the required trend line.

 The owner of a small company manufactures a product. Since he started the company, the number of units of the product he has sold is represented by the following time series:

Year Demand

1995 1996 1997 1998 1999 2000 2001 100 120 95 105 108 102 112

Find the trend line that describes the trend by using Semi-averages and forecast the demand for 2005

 Fit a trend line to the following data by the method of semi-average and forecast the sales for the year 2002.

Year
1993 1994 1995 1996 1997 1998 1999

Sales
102 105 114 110 108 116 112

 Measurement of Seasonal effects Seasonal variation is defined as repetitive and predictable movement around the trend line in one year or less. Seasonal effects on time series is essential for projection of past pattern into the future for short run predictions. Seasonal effects are measured in terms of an index. A seasonal index is an average that indicates the percentage deviation of actual values of the time series from a base value which excludes the short term seasonal influences.

 Methods:

 Method of simple average  Ratio-to trend method  Ratio-to-moving average  Link relatives method

 Ratio-to-moving average method:

It provides an index for seasonality that describes the degree of seasonal variation.

 Procedure: 1. Calculate the 4-quarter moving average for the time series (for quarterly data). Calculate 12 month moving average for the data ( for monthly data). 2. Compute centered moving averages. 3. Calculate the % of the actual value to the moving average value for each quarter in the time series. Ratio-to-moving average= ( Actual value/Moving average) *100

4. Arrange all the % of the actual to moving average values by quarter. 5. Calculate the modified mean by discarding the highest and lowest values for each quarter and average the remaining values. 6. Adjust the modified mean to 400 for quarterly data, 1200 for monthly data by multiplying each of the quarterly indices (monthly indices) by an adjusting component and is found by dividing the desired sum of the indices(400/1200) by the actual sum. This process brings to a total of 400 for quarterly data, 1200 for monthly data.

 The resort hotel wanted to establish the seasonal pattern of room demand by its clientele. Hotel management wants to improve customer service and is considering several plans to employ personnel during peak periods to achieve this goal. The following table contains the quarterly occupancy i.e the average number of guests during each quarter of the last 5 years.
Year 1991 1992 1993 1994 1995 I 1861 1921 1834 1837 2073 No.of guests per quarter II III 2203 2415 2343 2514 2154 2098 2025 2304 2414 2339 IV 1908 1986 1799 1965 1967

Solution

 De seasonalized occupancy ! Actual occupancy Seasonal index 100

 Suppose the hotel management estimates from a de-seasonalized trend line that the deseasonalized average occupancy for the 4 th quarter of the year 1996 as 2121. what will be the seasonalized average occupancy?

seasonalized occupancy ! seasonal occupancy deseasonalized occupancy v 100

 Method of Simple Averages: 1. Average the data by the years and months (quarters if quarterly are given). 2. Add the figures of each month (quarter) and obtain the average by dividing the monthly (quarterly) totals by the number of years. 3. Obtain an average of monthly (quarterly) averages by dividing the total of monthly(quarterly) averages by 12 (4). 4. Compute seasonal indices for different months (quarters) by expressing monthly averages (quarterly averages) as % of the grand average.

 The data below give the average quarterly prices of a commodity for four years. Calculate the seasonal indices.

u r arte e ar 8 8 8 8 . . 7. . .8 . . . . . 8 . . .

Solution

 Measurement of Cyclic Variation:

y-y Relative cyclical residual ! v 100 y

 A farmer s marketing cooperative wants to measure the variations in its member s wheat harvest over an year period. The following table shows the volume harvested in each of the years.

Year 1988 1989 1990 1991 1992 1993 1994 1995 Actual Bushels '0000s) 7.5 7.8 8.2 8.2 8.4 8.5 8.7 9.1

Solution