Business

Economics
Regression Analysis
Prof. Dr. Qais Aslam

Time Series Analysis • Time Series Refers to the values of a variable arraigned chronologically over time • The First Step is to plot past values of the variable in question on the vertical axis • Time Series attempts to forecast future values of the time series by examining past observations of the data only • The Assumption is that the time series data will continue to move in the future as it has behaved in the past • Most Time Series Data varies or fluctuates over time .

natural disasters. Secular Trends: Refers to long term increase or decrease in the data series 2.Reasons for Fluctuation in Time Series Sessional Variation s Random Influenc es Secul ar Trend s Cyclic Fluctuatio ns 1. Cyclic Fluctuation: are major expansion or contraction in most of the economic time series that seem to reoccur every few years 3.Note: The Total variation in the Time Series results from all four factors operating together . Regular or Random Influences: are the variations in the data series resulting from unexpected occurrences as wars. strikes or other unique occurrences •. Seasonal Fluctuations: Refer to regularly reoccurring fluctuations in the economic activity every year 4.

Trend Projection in the Time Series •  Simple form is the projection of the past trends by fitting a straight line through regression analysis: Where St is the value of the time series to be forecasted in time t is the estimated value of the time series or the constant of the regression in the base period t = 0 is the absolute amount of growth per Trend Line period is the time period to be forecasted in the .

is the time period to be forecasted in the time series analysis • While the assumption of a constant absolute amount of change per period may be appropriate in many cases. there are situations where a constant percentage change is more appropriate. Therefore the constant percentage growth rate model can be specified as: (2) • Where g is the constant percentage growth rate to be estimated • To estimate g from equation (1) we must first transform the time series data in Lanier in the logarithms (3) . is the estimated value of the time series or the constant of the regression in the base period t = 0. is the absolute amount of growth per period.Constant Percentage Growth Rate Model •  (1) Where St is the value of the time series to be forecasted in time t.

we simply find the average ratio by which the actual value of the time series differs from corresponding estimated trend values in each period under consideration during the3 long time period in the study • Very similar results can be obtained by the inclusion of Seasonal Dummy variables ) • Note: That the estimated coefficients for the dummy variables and the trend variables are all statistically significant at better than 1% level and that equation (3) explains 99% of the variation .Sessional Variation • To   adjust the trend forecasts for the sessional variations by the ratio-to-trend method.

we find the average ratio by which the actual value of the time series differs from the corresponding estimated trend value in each period of the long time period under study • Note: There is a much more sophisticated method of time series analysis called the Box. the value of the constant of the regression is taken as the base in the equation • To adjust the trend forecasts for the sessional variation by the ratio-to-trend method.• That the dummy variable is added to if positive and subtracted from if negative.Jenkins Technique which is much more complex than the above analysis technique (which is not the scope of this subject) .

The greater the number of periods used in the moving averages. the greater is the smoothing .Smoothing Techniques • Other methods of naïve forecasting are Smoothing Techniques: These predict future values of a time series on the basis of some averages of its past values only • Moving Averages: are simple techniques where the forecasted value of a time series in a given period of time is equal to the average value of the time series in a number of previous periods under study.

Root Mean Square Error (RMSE) • RMSE is calculated for each forecast and   utilize the moving average under the formula: (4) • Where is the actual value of the time series in a period t • is the forecasted value and • is the number of time periods or observations .

Exponential Smoothing • Exponential smoothing overcomes the criticism   made on average forecasting • With exponential smoothing.) (5) . the greater is the weightage given to the value of the time series in period t as opposed to the pervious periods. • The value of time series at period t or is assigned a weight () between o and 1 inclusive. the forecast for period t + 1 that is is a weightage average of the actual and forecasted values of time series in period t.. and the forecast for period t or is assigned the weightage between 1 . • Thus the value of the forecast of the time series in period t + 1 is: = + (1 .

we must also decide on the value of or the weightage assigned to • In general different values of are tried.) (5) • Note: The sum of the weightage equals 1 or + (1 . it is necessary to assign value to the initial forecast to get the analysis started • One way to do this is to let equal the mean value of the entire observed time.series data • Second. and the one that leads to the forecast with the Smallest Root . Therefore the technique is called Exponential Smoothing • Two decisions must be taken in order to use the above equation: • First.Mean – Square .•  = + (1 .Error (RMSE) is actually used in the forecast . • The forecast for period t can be shown to depend on the past values of the time series with weightages declining exponentially for values further into the past.) = 1 • While is calculated from the value of the time series and its forecasts for period t only.