This thesis focuses on the route choice behavior of car drivers (uni-modal networks). More precisely, we are interested in identifying which route a given traveler would take to go from one location to another. For the analysis of this problem we use discrete choice models and disaggregate revealed preferences data. Route choice models play an important role in many transport applications, for example, intelligent transport systems, GPS navigation and transportation planning. The route choice problem is particularly difficult to analyze because it involves the modeling of choice behavior in large transportation networks. Several issues need to be addressed in order to obtain an operational model. First, trip observations in their original format rarely correspond to link-by-link descriptions of chosen paths and they therefore need to be matched to the network representation used by the modeler. This involves data processing that can introduce bias and errors. Second, the actual alternatives considered by the travelers are unknown to the analyst. Since there is a large, possibly infinite, number of feasible paths in the network, individual specific choice sets of paths need to be defined. Third, alternatives are often highly correlated due to physical overlap between the paths (shared links). Models with flexible correlation structure are complex to specify and to estimate. Simple models are therefore often used in practice even tough the associated assumptions about correlation are violated. Fourth, most route choice models assume that the decision is performed pre-trip. Their application in a context where drivers receive real-time information about traffic conditions is questionable. In this thesis we address each of the aforementioned issues. First, we propose a general modeling scheme that reconciles network-free data with a network based model so that the data processing related to map-matching is not anymore necessary. The framework allows the estimation of any existing route choice model based on original trip observations that are described as sequences of locations. We illustrate the approach with a real dataset of reported long distance trips in Switzerland. Second, a new paradigm for choice set generation in particular and route choice modeling in general is presented. Instead of focusing on finding alternatives actually considered by travelers, we propose an approach where we focus on obtaining unbiased parameter estimates. We present a stochastic path generation algorithm based on an importance sampling approach and derive the corresponding sampling correction to be added to the path utilities in the route choice model. This new paradigm also has implications on the way to describe correlation among alternatives. We argue that the correlation should be based not only on the sampled alternatives but also on the general network topology. Estimation results based on synthetic data are presented which clearly show the strength of the approach. Third, we propose an approach to capture correlation that allows the modeler to control the trade-off between the simplicity of the model and the level of realism. The key concept capturing correlation is called a subnetwork. The importance and originality of this approach lie in the possibility to capture the most important correlation without considerably increasing the model complexity. This makes it suitable for a wide spectrum of applications, namely involving large-scale networks. We illustrate the model with a GPS dataset collected in the Swedish city of Borlänge. The final contribution of this thesis concerns adaptive route choice modeling in stochastic and time-dependent networks, as opposed to the static network setting assumed in existing models. Optimal adaptive routing problems have been studied in the literature but the estimation of such choice models based on disaggregate revealed preference data is a new area. We propose an estimator for a routing policy choice model and use synthetic data for illustration. Given the uncertainty related to travel times and traffic conditions in transportation networks, we believe that adaptive route choice modeling is an important direction for future research. To summarize, this thesis addresses issues related to data processing (network-free data approach), algorithms for choice set generation (sampling of alternatives) and models (subnetwork approach and adaptive route choice model). Moreover, we use real applications (Borlänge GPS dataset and reported trips in Switzerland) to illustrate the models and algorithms.