A model for estimating program size, before coding of the program begins, on the basis of the complexity of the software is presented. The complexity of the software is computed from program logic charts with the Halstead metrics. The relationships between the complexity of the software and program size are then statistically analyzed by taking hypothesized environmental factors, such as program categories, into consideration. Results of an experiment show that, of the Halstead metrics, N/sub 1/ is the most significant and reasonable as a metric for computing the complexity of the software from program logic charts. The results also show that the model based on this metric is more reliable than the conventional size estimation method.>

Many simple software errors are found in earlier software test phases.The ratio of complex errors to simple errors gradually increases with continual testing.This paper describes a software reliability model called the Error Complexity Model.In this model, errors are classified by error complexity which is a measure of error detectability.The number of remaining software errors is estimated from the ratio of complex to simple errors and the number of discovered errors.New criteria for error complexity classification are proposed.The model is evaluated and compared with existing models using actual error data.