Predicting seismic retrofit construction cost for buildings with framed structures using multilinear regression analysis

Attempts to predict construction cost represent a problem of continual concern and interest to both practitioners and researchers. Such an attempt is presented here for the specific challenge of cost prediction when undertaking seismic retrofitting of existing structures. Using multilinear regression analysis, 14 independent variables were analyzed to develop parametric models for predicting the retrofit net construction cost (RNCC). Half of these variables have never previously been studied in the literature. The required data for this study were collected from 158 earthquake-prone public schools in Iran, each having a framed structure. The backward elimination (BE) regression technique was used to identify any variables that made a statistically significant contribution to the RNCC. The suitability of the BE technique for this identification was examined and demonstrated using a number of model-selection criteria. Rather surprisingly, building age and compliance with the earliest practiced seismic design code were found to be insignificant predictors of the RNCC. As reflected by the BE technique, the significant predictors were building total plan area, number of stories, structural type, seismicity, soil type, weight, and plan irregularity. The causal analysis performed between the RNCC and these variables showed that the first two variables have the greatest influence on the determination of the RNCC. The primary contribution to the construction industry is the introduction of a simple double-log cost-area model for predicting seismic retrofit construction cost. The introduced model enables engineering consultants, managers, and policy makers to simply predict this cost at the early planning and budgeting stage of seismic retrofit projects.