10DBMC International Conférence On Durability of Building Materials and Components LYON [France] 17-20 April 2005 Estimating survival functions of building stocks P.E. Bradley, C. Ferrara, N. Kohler Institut für Industrielle Bauproduktion, Universität Karlsruhe Englerstraße 7, D-76128 Karlsruhe, Germany bradley@ifib.uni-karlsruhe.de TT5-63 ABSTRACT When dealing with building stocks, one is often faced with the quite hard problem of obtaining sufficient information for studying its behaviour over time. In particular, a survival analysis of buildings in large building stocks seems an impossible task, if one wants to include all available data on construction and demolition of individual buildings in the stock. Random sampling seems also not feasible, because a priori information on the number of demolished buildings is needed. In this paper, an example-based methodology is developped for estimating survival functions of large building stocks for which the lifetime data is not very easily acquired, e.g. for townships whose corresponding data has to be extracted from paper archives. By considering demolished as well as non-demolished buildings, techniques from censored and truncated data are applied: the event considered being demolition, non-demolished buildings are then right-censored, thus leading one to calculate the Kaplan-Meier estimator for the survival function. This methodology is applied to a German middle-sized small town, where, due to the data acquisition problem, the right-censoring itself is estimated by a random sample from a cadastral plan, and a complete inventory count is performed for the demolished part of the building stock. The resulting estimates for the survival functions are compared for various substocks generated by stratification (residential and non-residential buildings by age classes). In a further step, parametric survival functions are fitted in order to allow prediction of the behaviour of a building stock. The example building stock was studied in the project "Validierung eines integrierten, dynamischen Modells des deutschen Gebäudebestandes" (Validating an integrated, dynamic model of the German building stock) funded by the Deutsche Forschungsgemeinschaft 1999- 2003, and aiming at developping methods for collecting and analysing large-scale building stock related data (such as for townships). KEYWORDS Building stock, right censoring, survival analysis, Kaplan-Meier estimator.