Objectives: This study explains why frequency polygons for US birthweights in 100-g weight classes appear spiky compared with their European counterparts.
Methods: A probability model is used to describe how unit conversion can induce misclassification. Birthweights from the United States and Norway are used to illustrate that misclassification operates in grouped US data.
Results: Spikiness represents misclassification that arises when measured birthweights are rounded to the nearer ounce, converted to grams, and then grouped into weight classes. Misclassification is ameliorated, not eliminated, with 200-g weight classes.
Conclusions: Possible biases from misclassification should be carefully evaluated when fitting statistical models to grouped US birthweights.