Correlations are concerned for modeling of CMOS devices and circuits variability when using back propagation of variations (BPV) methodology in the paper. Strong reverse correlations are observed and investigated between variation parameters, particularly for threshold voltage (V-T), total gate capacitance (C-gg) and trans-conductance (g(m)), on gate voltage dependence due to random dopant fluctuation (RDF) in nanometer MOSFETs. These correlations are verified both in theoretical and simulation approaches. Based on these correlations, a simple and accurate analytical model for capturing f(T) variability is proposed. The model is in good agreement with HSPICE Monte Carlo simulations in different design decisions such as effective width length ratios, source voltages and doping concentrations. Results show the estimation errors are not more than -2.33% and 1.30% for HMOS and PMOS, respectively. Furthermore, our analysis of the correlation and analytical formula are still effective for the continued scaling CMOS technology. (C) 2014 Elsevier Ltd. All rights reserved.