Genotype by environment interaction is one of the most important subjects in crop improvement, so that different statistical measures have been proposed to estimate this interaction. In the present study, nine varieties and hybrids of spring rapeseed were evaluated in five warm dryland areas including Sarpol-e-Zahab, Gonbad, Gachsaran, Pol-e-Dukhtar and Masjedsoliman in two growing seasons 2002-03 and 2003-04. Varietal trials were conducted in a RCBD with three replications in rain-fed conditions. The analysis of stability of the varieties was based on nonparametric statistics. Results showed that based on four nonparametric measures S_i ⁽¹⁾, S_i ⁽²⁾, S_i ⁽³⁾ and S_i ⁽⁶⁾, three varieties including Regent, Goliath and Global were selected through their seed yield were less than the overall mean. Use of three nonparametric measures NP_i ⁽²⁾, NP_i ⁽³⁾ and NP_i ⁽⁴⁾ led to the selection of two varieties Goliath and Regent with yield less than the overall mean. NP_i⁽¹⁾ recognized two varieties Kristina and Goliath as yield stable genotypes which yield of the first one was  more than overall mean and the second one was less than that. All eight nonparametric measures showed negative or non-significant correlation with yield. RI and L_i measurements selected four genotypes, including Hyola401 and Quinta with yield more than overall mean and Global and Goliath with yield less than the overall mean. Calculation of rank sum measure revealed that selection of stable genotypes based on this measure led to genotypes with a yield higher than the overall mean. In this study rank sum measure was the best method among different non-parametric measures. On the base of this measure Quinta, Shiralee and Option500 selected as seed yield stable genotypes.