Disassembly leveling and lot-sizing is an integrated problem of determining the depth of disassembling a product and the resulting amounts of disassembling the product and its subassemblies to satisfy component demands. This study addresses two stochastic versions of the problem for multiple product types with uncertain component demands: a basic problem and its extension with parts commonality. The objective is to minimize the sum of expected setup and operation costs. For the basic problem without parts commonality, an optimal sample average approximation algorithm is proposed and then illustrated with a numerical example after formulating it as a stochastic integer programing model. Also, due to the complexity of the extended problem, sample average approximation based heuristics are proposed and computational results are reported. In particular, the results show that the best heuristic gives near optimal solutions for small sized test instances.