Abstract Height/age curves of dominant trees in a stand can be predicted using models containing random stand and tree parameters. If the population mean curve is estimated from the data where stand-specific parameters are correlated with stand age (e.g., because old stands are growing on poor sites), this correlation may cause biased estimates. The same bias problem occurs in traditional site index methods. It is shown how ordinary least squares estimates of tree specific parameters can be used for computing unbiased estimates of the population mean parameters and for estimating the relationship between stand parameters and stand age. The application of the estimated model several years after the data collection time is problematic because it should be known how the population changes over time. For. Sci. 40(4):715-731.