• Overview
• Files
• Data Sets
• Seriations
• Features
• Artifacts
• Assemblages
• Relations
• Data
• Seriate
• Results
• Graphs
• View
• Window
• Help
  • Glossary

Confidence Intervals

Confidence intervals are available on the Values Graph. They are only available if you are using frequencies (Use Frequencies on the Seriations table), and if the numerical entries in your data are all integer. The menu also has an option for setting the confidence level.

Confidence intervals are intended to be used with frequency seriation. For a particular item (artifact or assemblage, but normally an assemblage with frequency seriation), each of the input data values must be blank, zero or another integer. If n is the sum of the data values for all features (with frequency seriation a feature is usually a style) for the item, and k is the value of a particular style, then f = k / n is the frequency for that assemblage and style. The frequencies graph displays these frequencies. Given the data values for the assemblage, it is possible to calculate a confidence interval for each style. If m is the number of styles present in a particular assemblage, then the confidence interval is calculated using a beta prior distribution with parameters (1, m - 1) and a beta posterior distribution with parameters (k + 1, n - k + m - 1). This is directly affected by how the Blanks and Zeroes parameters and the Earlier and Later parameters are set. For example, if Blanks are set to Present & Zero, then a blank will increase the value of m, but if Blanks are set to Absent & Zero, then a blank will not affect the value of m.