Recently, Onatski (2009) provided a formal justication for the use of the scree test in determining the number of factors in large factor models. We derive a criterion for the determination of the number of factors, that is associated with Onatski's test statistic but does not rely on frequency domain analysis. Monte Carlo experiments compare the nite-sample properties of our criterion to Bai and Ng (2007), Hallin and Liska (2007) and Onatski (2009), and show that our criterion is superior. An application with the U.S. macroeconomic data set of Stock and Watson (2005) illustrates our procedure.