In the simulation of nuclear plant behavior through system codes, there are often uncertainties associated with the large number of model parameters required as code inputs. The use of the Taguchi method is investigated for the importance ranking of uncertainties when a single metric is used to characterize system performance. The proposed procedure is illustrated on a simplified boiling water reactor (BWR) model to determine the dominant parameters affecting the maximum limit cycle amplitude (MLCA) in BWRs. A reduced-order BWR model is used for the analysis. A regression model is also generated to predict the MLCA as a function of the parameter values in their assumed uncertainty regions. The results indicate that (a) 7 out of the 11 parameters (factors) under consideration have a significant impact on the MLCA, (b) a linear regression model can be constructed to predict the MLCA with 88% confidence, (c) higher-order effects of the control factors are negligible, and, (d) cross effects between the factors are negligible compared to their individual effects. The results also indicate that the use of the Taguchi method leads to a 99.4% reduction in the computational effort over a full factorial experiment design. The use of the Taguchi method is not proposed to replace the well-established conventional methods for sensitivity and uncertainty analysis but rather to assist them in the selection of the parameters that may require more detailed analysis.