Counts, shares, margins

◈ 9 cards

table(orders$prov) is AB 2 BC 3 ON 4 QC 3; round(prop.table(…) * 100, 1) is 16.7 25.0 33.3 25.0; sort(table(x), decreasing = TRUE) puts ON first. table(orders$prov, orders$channel) is the pivot Rows × Columns with Count; addmargins() adds the Sum row and column.

Counting categories

Module 4 counted provinces with COUNTIF, four times; Module 7 did it once with a pivot. table() is the one-call R version:

> table(orders$prov)

AB BC ON QC 
 2  3  4  3

Level names on the first line, counts on the second, alphabetical. Missing values are not counted unless you ask (useNA = "ifany").

Worked example — from counts to shares

A count answers "how many"; a share answers "how much of the whole". prop.table() divides a table by its total:

> prop.table(table(orders$prov))

       AB        BC        ON        QC 
0.1666667 0.2500000 0.3333333 0.2500000 
> round(prop.table(table(orders$prov)) * 100, 1)

  AB   BC   ON   QC 
16.7 25.0 33.3 25.0

ON is a third of orders. The * 100 and round(…, 1) turn a proportion into a percentage to one decimal — the pivot's Show Values As → % of Grand Total, and the same 33.3 % Module 7 read off it.

A table sorts like a vector:

> sort(table(orders$prov), decreasing = TRUE)

ON BC QC AB 
 4  3  3  2

Most frequent first — the order a bar chart of counts should use.

Two variables: the pivot in one call

table() with two columns is Rows × Columns with Count. The first argument goes in the rows:

> table(orders$prov, orders$channel)
    
     Online Store
  AB      1     1
  BC      2     1
  ON      2     2
  QC      1     2

ON placed 2 online and 2 in-store orders. addmargins() adds the totals the pivot shows by default:

> addmargins(table(orders$prov, orders$channel))
     
      Online Store Sum
  AB       1     1   2
  BC       2     1   3
  ON       2     2   4
  QC       1     2   3
  Sum      6     6  12

The Sum column is table(orders$prov) again; the Sum row is table(orders$channel) — 6 and 6. prop.table(t, 1) gives row shares (each province split by channel) and prop.table(t, 2) column shares, matching % of Row Total and % of Column Total.

Type the share and the margins, then reproduce the shares in Python

The code block counts with a dictionary and prints each province's share of 12 to one decimal, sorted — the same four numbers.

CRISP-DM: counts and shares are data understanding → explore data.