Over the years, accurate prediction of passenger and freight transport movements became essential to identify suitable policies that solve the increasing challenges caused by these movements. Freight demand can be modelled either as Freight Trip Generation (FTG) or Freight Generation (FG). While modelling commodities in tonnage, pallets, or deliveries is FG, FTG models freight vehicle trips. Many studies model FTG over FG and employ either category analysis or Ordinary Least Squares (OLS) model, which ignore FTG characteristics. Count models that consider most FTG characteristics are rarely employed for modelling. Also, most studies segment FTG by establishment type. These studies consider only freight trips by trucks or assume that all freight trips are equivalent to a truck trip. However, freight in several large cities is increasingly being moved by smaller vehicles and thus segmentation by establishment category may underestimate the freight trips generated and its effects. This calls for modelling FTG segmented by vehicle type. FTG models also rarely consider spatial interactions. Besides, the econometric model that provides the best fit for FTG is not identified. Freight data employed in FTG and FG models is predominantly obtained through Establishment-Based Freight Surveys (EBFS). EBFS suffer from the unit and item nonresponse. While the consequences of unit nonresponse and the possible measures to reduce the effect of unit nonresponse on the sample's representativeness of the population are studied, item nonresponse is not studied. The present study aims to fill these gaps. In addition to providing FG and FTG rates, it also develops two sets of models for FTG segmented by vehicle type, one without and the other considering spatial interactions. Each set of models developed include OLS, Seemingly Unrelated Regression (SUR), Poisson, Negative Binomial (NB), Zero-inflated Poisson (ZIP) and Zero-inflated Negative Binomial (ZINB) models. The results show that the spatial models provided better fit over non-spatial models. The spatial ZINB model provides the best fit for FTG in most cases. The study also proposes development of FG models segmented by establishment category and a combined model for all establishments. Further, it also proposes to analyse item nonresponse in EBFS to identify measures to attain a more representative sample of the population.