The time scale of failures need not be chronological, if a failure occurs according to the accumulation of damages from its usage or exposure to some risk. The true time scale, if there could be, need not show the best fit among time scale candidates to the observed field reliability data. A semiparametric estimator of the time scale which combines multivariate lifetime data has been proposed in the literature. It does not require the specification of the lifetime distribution beforehand on the time scale also to be estimated. We first show a set of simulation results for investigation of the properties of time scale models and the sample properties of their estimates. Then we show another result on the performance under the time scale misspecifications. Finally the estimator and the time scale functions are applied to a problem of finding a suitable time scale for field reliability data. We show that the results are useful even in cases when there exists prior knowledge about the suitable time scale.