An empirical study of software productivity

Measuring and predicting software productivity: A systematic map and review

“…The studies approached the problem of identifying productivity in a similar manner by evaluating the explanatory power of the model. It is shown that all models provide a high explanatory power for the variance of productivity in their environments, which is true for the studies by Foulds et al [46], Morasca and Russo [45], Maxwell et al [17] and Jeffery [43] (see Table 4). The studies were based on the ideas in regression analysis.…”

“…[5]). A multivariate logistic regression model was evaluated by Morasca and Russo [45]. In their logistic regression the dependent variable (Y) was ordinal/nominal, and discretized into classes.…”

“…Function points, experience, availability of development time are significant predictors [45]. Lines of code and maximum staff as dominating factors, and average team attitude to the project and average procedural language experience with minor impact on the productivity [43].…”

“…Regression applied to a variety of domains: Web projects [37], information systems [43,15,[44][45][46], and military/space/others [17] Different forms of regression employed leads to few studies per regression approach Overall high scores on evidence (all articles have scores greater than five, see Table 17) All studies have been conducted in an industrial context (evaluation research) Four of the studies are from before 1995 (relatively old data sets), see Table 4 No standard set of variables that is considered in every study hinders in arriving at a general understanding of predictors (see Table 4) Primary criterion for evaluation is limited to explanatory power (see Table 18)…”

“…[46,45,43]). The studies approached the problem of identifying productivity in a similar manner by evaluating the explanatory power of the model.…”