© 2016 American Physical Society. A systematic study of the pseudorapidity dependence of elliptic flow parameter using transport models (e.g., a multiphase transport model, AMPT, and ultrarelativistic quantum molecular dynamics, UrQMD) has been presented. We have observed that while at mid-pseudorapidity the elliptic flow measured using the event-plane method differs significantly from that measured by actual reaction plane method, both the event-plane and reaction-plane methods give the same elliptic flow for far forward and backward pseudorapidity. This indicates that the magnitude of measured v2 around midrapidity strongly depends on the analysis method. Therefore, one should use the same procedure (as used in data analysis) in model calculations while comparing model results and experimental data. We find the shape of v2(η) measured by the PHOBOS experiment is not reproduced by using actual v2 (i.e., measured with respect to the reaction plane) from AMPT and UrQMD models. The shape and magnitude of measured v2(η) can be explained by the AMPT model with string-melting mode only if one uses the same procedure as used in data analysis. Magnitude of elliptic flow can be reproduced for all pseudorapidity range by taking the parton-parton interaction cross section to be 3 mb at sNN=62.4 and 200 GeV. This implies that the partonic interactions are necessary to reproduce data at sNN=62.4 and 200 GeV and the strength of partonic interactions at far forward and backward rapidity is as strong as at midrapidity. Both UrQMD and AMPT with default mode fail to explain the data.