Suggested technique of Matched Paired test in fitted (Linear & Non-Linear) models and measuring their fitness by Auto regression ANOVA in estimating of the Predication's accuracy in time series

Abstract

The research aimed to estimates the goodness – of – fit tests for ( Linear & Non-Linear ) Models in the analysis of variance for the Autoregression in time series as an conventional ( studied ) method and companied the fitness results by the results of matched paired testing in the related groups as a suggested method , In addition to that , priority of the ordered goodness – of – fit according to several assuming with application contents the product quantities of Iraqi cigarettes through the period ( 2000 – 2009 ) in addition to the data of the time series for the account of sales of that quantities by supposed an Autoregression model of order one ( i.e. AR (1)) . The results showed similarity in the testing goodness – of – fit for all supposed models and for the two series at the level of significant ( P < 0.05 ) in a despite of differences in their priority ordered through applying the two methods. That indicated regression or Autoregression ANOVA would not standup under close scouting for testing goodness – of – fit method which had been mentioned by several literatures .