ECON 2B03 Lecture Notes - Lecture 2: Observational Error, Frequency Distribution, Frequency (Statistics)
Document Summary
When presenting data either in the form of tables of graphs, it is often desirable to first slit the data into groups/classes. How we split the data often depends on the data types. Data types: categorical (nominal or ordinal, quantitative (discrete or continuous) Regardless of the data type, data classes should be: collectively exhaustive (must exchange all logical possibilities for classifying available data, mutually exclusive (must not overlap or have data in common) One immediately confronts the issue of how many classes to create. Desirable class number: should fit data type, often recommended: between 5 and 20, sturgess"s rule: desirable number of classes = k, an integer, where k is the integer closest to (use standard rules for rounding) Where n is the sample size, and where log10n is the power to which the base (10) is raised to yield n.