Analysis of intersection crashes is a significant area in traffic safety research. This study contributes to the area by identifying traffic-geometric characteristics and driver demographics that affect different types of crashes at signalized intersections. A simple methodology to estimate crash frequency at intersections based on the size of the intersection is also developed herein. First phase of this thesis used the crash frequency data from 1,335 signalized intersections obtained from six jurisdictions in Florida, namely, Brevard, Seminole, Dade, Orange, and Hillsborough Counties and the City of Orlando. Using these data a simple methodology has been developed to identify the expected number of crashes by type and severity at signalized intersections. Intersection size, based on the total number of lanes, was used as a factor that was simple to identify and a representative of many geometric and traffic characteristics of an intersection. The results from the analysis showed that crash frequency generally increased with the increased size of intersections but the rates of increase differed for different intersection types (i.e., Four-legged intersection with both streets two-way, Four-legged intersection with at least one street one-way, and T-intersections). The results also showed that the dominant type of crashes differed at these intersection types and severity of crashes was higher at the intersections with more conflict points and larger differential in speed limits between major and minor roads. The analysis may potentially be useful for traffic engineers for evaluating safety at signalized intersections in a simple and efficient manner. The findings in this analysis provide strong evidence that the patterns of crashes by type and severity vary with the size and type of intersections. Thus, in future analysis of crashes at intersections, the size and type of intersections should be considered to account for the effects of intersection characteristics on crash frequency. In the second phase, data (crash and intersection characteristics) obtained from individual jurisdictions are linked to the Department of Highway Safety and Motor Vehicles (DHSMV) database to include characteristics of the at-fault drivers involved in crashes. These crashes are analyzed using contingency tables and binary logistic regression models. This study categorizes crashes into three major types based on relative initial movement direction of the involved vehicles. These crash types are, 1) Initial movement in same direction (IMSD) crashes. This crash type includes rear end and sideswipe crashes because the involved vehicles for these crashes would be traveling in the same direction prior to the crash. 2) Initial movement in opposite direction (IMOD) crashes comprising left-turn and head on crashes. 3) Initial movement in perpendicular direction (IMPD) crashes, which include angle and right-turn crashes. Vehicles involved in these crashes would be traveling on different roadways that constitute the intersection. Using the crash, intersection, and at-fault driver characteristics for all crashes as inputs, three logistic regression models are developed. In the logistic regression analyses total number of through lanes at an intersection is used as a surrogate measure to AADT per lane and also intersection type is introduced as a 'predictor' of crash type. The binary logistic regression analyses indicated, among other results, that at intersections with one-way roads, adverse weather conditions, older drivers and/or female drivers increase the likelihood of being at-fault at IMOD crashes. Similar factors associated with other groups of crashes (i.e., IMSD and IMPD) are also identified. These findings from the study may be used to develop specialized training programs by zooming in onto problematic intersections/maneuvers.