modeling problems involving bivariate dependent data is very important in many areas, such as finance, actuary, reliability and survival analysis. in the literature, some copula models have been widely used to modelling dependent multivariate distributions, among which stands out the archimedean copula class. this paper presents a methodology to select from some archimedean copula models the one that fits the best to a dependent dataset, using goodness-of-fit plots, q-q plots and cramér-von mises goodness-of-fit test. we illustrated the methodology using simulated data and data from insurance claims. the results showed that the data insurance fits to a bivariate model based on the so called frank's copula with lognormal marginals.