Techniques are presented for estimating the magnitude and frequency of flood discharges on rural, unregulated streams in New York, excluding Long Island. Peak-discharge-frequency data and basin characteristics from 388 streamflow-gaging stations in New York and adjacent states were used to develop multiple linear regression equations for flood discharges with recurrence intervals ranging from 1.25 to 500 years. A generalized least-squares (GLS) procedure was used to develop the regression equations. Separate sets of equations were developed for each of six hydrologic regions of New York; standard errors of prediction range from 14 to 43 percent. Statistically significant explanatory variables in the regression equations include drainage area, main-channel slope, percent basin storage, mean annual precipitation, percent forested area, a basin lag factor, a ratio of main-channel slope to basin slope, mean annual runoff, maximum snow depth, and percentage of basin above 1,200 feet. Drainage areas for the 388 sites used in the analyses ranged from 0.41 to 4,773 square miles. Methods of computing flood discharges from the regression equations differ, depending on whether the estimate is for a gaged or ungaged basin, and whether the basin crosses hydrologic-region or state boundaries. Examples of computations are included. Discharge-frequency estimates for an additional 122 streamflow-gaging stations with significant regulation or urbanization (including Long Island) are also included as at-site estimates. Basin characteristics, log-Pearson Type III statistics, and regression and weighted estimates of the discharge-frequency relations are tabulated for the streamflow-gaging stations used in the regression analyses. Sensitivity analyses showed that mean-annual precipitation, drainage area, mean annual runoff, and maximum snow depth are the variables to which computed discharges are most sensitive in the regression equations. Included with the report is a DVD that provides computation procedures and geographic information system spatial datasets to compute basin characteristics used in the regional regression equations and flood-frequency estimates at a specified location on a stream.