Statistical and econometric modeling is used to analyze the dynamics of generators? bidding behavior and resulting market prices in the NYISO day-ahead market during the summer of 2001 at three levels. First, at the generator level, generators? offered capacity is modeled by a two-stage sample selection process. The grouping choices of generators are estimated in the first stage. In the second stage, generators? offered capacity is estimated conditional on their grouping choices. Second, at the group level, we estimate the offer curves of each group. Third, at the total supply level, the aggregate supply curve is simulated by aggregating the offer curves of bidding groups. The estimated market prices based on the simulated supply curve can replicate most volatility of actual market prices. We find that demand conditions affect generators? bidding behavior and thus market prices significantly by applying our model under different demand assumptions.