Repeat sales price estimators are designed to infer price indexes of infrequently sold and unstandardized assets, such as houses, based only on changes in prices of those individual assets that are observed to be sold twice. Repeat sales price estimators are proposed here that are arithmetic, and either value-weighted or equally-weighted. Moreover, variants are proposed that are interval-weighted, i.e., that correct for a form of heteroskedasticity, and that include additional regressors representing changes in hedonic variables. Some of these methods are applied to data on house prices in Atlanta, Chicago, Dallas and San Francisco 1970–1986.