Forecasting

Statistical Techniques in Forecasting

Statistical Techniques in Forecasting

Time Series Plot of IOP

Time Series Plot of UK manufacturing industries shows continuous variation in the output level. Historical data for output shows an inclining trend over the period from Year 1980 to Year 2005. Time series plot of the data is given below:

Trend Curve Selection

In order to assess the trend of the given output, multiple approaches can be applied depending upon the future prospects. Since the data given involves time factor which includes seasonal and dimensional changes in productivity level. Autoregressive Integrated Moving Average Model ARIMA considers equalize variances for the data. It also requires stationary series data which makes the expected value of series independent of time (Armstrong 2001, p. 57).

In some situations, non-stationary data can be converted to stationary data by removing the seasonal differences and incorporating moving average trend through ARIMA modelling technique (Creswell 2009, p. 92). However, in that case Exponential Smoothing of the data for managing the seasonality component and sequential trend would need to be applied. Trend cycle component increases the variation between data if seasonal trend remain unadjusted. This requires normalizing the irregular or random component factors to enhance the prediction of sequential output for the given case (Balnaves 2007, p. 156). Increase variation between data figures will require moving regression trend curve to assess the effectiveness of forecasted value and actual values against forecasted ones.

Considering the impact of seasonality and trend component for the given case, Expert Modeler will provide the best fitting model for analysing the future trend. Expert Modeler transforms the data using square root or log transformation in order to have statistically significant relationship with dependent series. It involves both ARIMA and Exponential Smoothing of data, which makes this trend curve fit for forecasting the output of manufacturing industries (Murugan 2007, p. 104). It excludes the trend of seasonality by identifying the errors component associated with the seasonal time. It provides adjusted seasonal series for the data that increases the probability to predict the seasonal factor component effectively; providing significant value in forecasting even in uncertain situations. Error identified in this modelling technique can be adjusted in forecasted value to enhance the accuracy of forecasted value.

Expert Modeler curve help analyze the current sales figure and set a trend of the industry adjusted with the change in output. Fitted value derived from the seasonal factor adjustment and trend cycle provides an advantage to have exponential trend curve for the forecasted value (Armstrong 2001, p. 59).

Quarter-Wise Forecast for Year 2006 & 2007

Forecasted Data obtained by applying Expert Modeler and adjustment of error in IOP Fitted reveals the following data:

Stationary R-Squared of IOP Fitted 0.809 appeared significant at 95%confidence interval which shows that future trend in the output will be as given in the trend forecast table.

Model Statistics

Model

Stationary R-squared

Statistics

Sig.

IOP Model

0.242

17.580

0.285

IOP Fitted

0.809

42.930

0.000

Although significance of IOP Model is comparatively higher however trend exhibited in the output is comparatively similar to IOP fitted values that involves adjustment of variance and error in forecasting ...