Regression analysis is explored for the multivariate data of two field experiments conducted with 7 sunflower genotypes, GUJ-SUN-1, MSFH-8, MSFH-17, KBSH-1, JWALA, PAC-36 and MORDEN, grown during kharif 1997 and 1998 in a dryland alfisol. Observations on 8 physiological traits, stomatal conductance, photosynthesis, leaf nitrogen, stem nitrogen, leaf weight, stem weight, leaf number and leaf area, as observed on 30, 45 and 60 days after sowing were used in the analysis for selecting the best genotype which is consistent in plant growth and has an efficient performance apart from identifying the dominant traits. Regression functions for predicting physiological traits have been calibrated by regressing the observations of a trait on 60 days after sowing through its observations on 30 and 45 days after sowing. The genotypes have been assessed based on the significance of traits, predictability and estimate of error derived for each trait. The analysis indicated the dominance of photosynthesis, stomatal conductance and stem nitrogen in kharif 1997 and stomatal conductance, photosynthesis and leaf nitrogen in kharif 1998 for the growth of sunflower. The analysis also indicated the superiority of PAC-36 in kharif 1997 and JWALA in kharif 1998 having 8 and 7 significant regression coefficients of traits, respectively.