00000ctm a22000004a 4500 UP-99796217611211924 Buklod 20140423131446.0 m |o d | ta 140423s2014 xx d r |||| u| (iLib)UPD-00214608061 DSTC rda eng DMLUC LG 995 2014 S8 T63 Tobias, Angelita P. author. A comparison of parametric and nonparametric estimation of multiple-input transfer function models Angelita P. Tobias. Quezon City School of Statistics , University of the Philippines Diliman 2014. 64 leaves illustrations 28 cm text rdacontent unmediated rdamedia volume rdacarrier Thesis (M.S.)--University of the Philippines Diliman. Abstract (Multi-input transfer function model is postulated as a nonparametric regression model. Issues and limitations of the parametric transfer function such as nonlinearity, correlated inputs, misspecification error present in the output series and short time series are addressed by the nonparametric model. The parametric transfer function is estimated via Box and Jenkins method while the nonparametric model is estimated through the backfitting algorithm. Simulation studies show that the nonparametric procedure is robust on the number of inputs, correlation among input series, time series length, form of transfer function, and misspecification errors. Moreover, the predictive ability for both procedure deteriorates for correlated inputs and is comparable. However, the nonparametric model becomes superior over the parametric model when there is misspecification). Multiple-input transfer function. Backfitting algorithm. Generalized additive models (GAM). FI UP UPD DARCHIVES LG 995 2014 S8 T63 UPD DSTC LG 995 2014 S8 T63 Thesis