Mathematics and Statistics · Ch 13 — Bivariate Frequency Distribution and Chi-Square Statistic
Marginal and Conditional Frequency Distributions
Marginal and Conditional Frequency Distributions
A two-way table actually carries three kinds of distribution inside it. The cell frequencies describe the two variables jointly; but by looking only at the totals, or only at a single row or column, two further, one-variable distributions can be read straight off the same table.
Marginal Frequency Distribution
The marginal distribution of is the set of row totals — the frequency distribution of on its own, obtained by adding across each row (i.e. ignoring entirely). The marginal distribution of is the set of column totals — the distribution of on its own. They are called marginal because they are written in the margins (the total row and total column) of the table.
Because a marginal distribution is just an ordinary single-variable frequency distribution, every earlier tool applies to it unchanged. In particular the mean of from its marginal distribution is
where is each -value's row total, and is found the same way from the column totals.
Conditional Frequency Distribution
A conditional distribution fixes one variable and looks at how the other is distributed within that fixed slice. The conditional distribution of given a particular value of is simply the frequencies in that one row; the conditional distribution of given a particular value of is the frequencies in that one column. Dividing each such frequency by that row's (or column's) total converts it to relative (proportional) form. …
The single-variable frequency distribution of one variable obtained from a two-way table's totals: the row totals give the marginal distribution of , the column totals give t …
The distribution of one variable within a fixed value of the other: the frequencies in a single row give the distribution of for that fixed ; the frequencies in a single column give the dist …