Stream flows calculated with the TR-55 method and obtained from USGS empirical data were compared for twenty stream gauges and their corresponding watersheds in New York State. The overall differences between the distributions of the two methods were measured using the Kolmogorov-Smirnov statistic, and 1, 10, and 100 year return period percent differences in flows. The three variables regressed against were average curve number, latitude and drainage area. Multiple linear regression, and lasso and ridge regression showed that none of the predictor variables had significant influence on the difference between modeled and the measured values. Area and latitude have higher correlations with the raw flows for the three return period storms than curve number when analyzed individually. This was not seen in multiple regression and may be representative of a curve number influence on the discrepancy between model and empirical data.