Definition
You stand by the road for ten minutes and count vehicles: car, bike, bike, auto, car, bus, bike… How do you turn that stream into something you can read at a glance?
A tally is a quick way of counting by hand, making one stroke for each item and crossing every fifth stroke so the marks fall into groups of five. A frequency table lists each category with its count, called its frequency, so the data can be read and compared easily.
In simple words
When a teacher counts votes for class monitor on the blackboard, they draw a line for each vote and cross every fifth one. At the end they count the bundles of five. That is tallying. Writing the final totals neatly in two columns, name and votes, makes a frequency table.
Why it matters
Counting is the first step of almost every data analysis. A frequency table turns a jumble of observations into a clear summary, and it is the basis for bar charts, averages and many AI datasets.
Key terms
- Tally mark
- A stroke made to count one item, grouped in fives.
- Frequency
- How many times a value or category occurs.
- Frequency table
- A table listing each category with its frequency.
- Category
- One group into which data is sorted, such as "bus" or "car".
- Total
- The sum of all frequencies, which equals the number of observations.
- Relative frequency
- A frequency shown as a share of the total, such as 12 out of 40, or 30 per cent.
Explained in detail
Why groups of five
People can recognise a group of up to about four objects at a glance without counting, but beyond that we start to count one by one and make mistakes. Grouping strokes in fives, four upright and one across, lets you count quickly in fives: five, ten, fifteen. The method is very old, and tally sticks with notches have been used to record counts for thousands of years.
Building a frequency table
- Decide the categories before you start, including "other".
- Make one tally mark in the right row for each observation.
- When you finish, count the tallies in each row: that is the frequency.
- Add all the frequencies. The total should equal the number of observations.
- If useful, add a column showing each frequency as a share of the total.
| Vehicle | Tally | Frequency | Share of total |
|---|---|---|---|
| Motorcycle | [5][5][5] || | 17 | About 43 per cent |
| Car | [5][5] | | 11 | About 28 per cent |
| Auto-rickshaw | [5] ||| | 8 | 20 per cent |
| Bus | ||| | 3 | About 8 per cent |
| Other | | | 1 | About 3 per cent |
| Total | 40 | 100 per cent |
Reading the table
The table shows at once that motorcycles are the most common vehicle, and that they make up more than four in ten of all vehicles counted. Showing each count as a share of the total makes it possible to compare with another road where you counted for a different length of time: 17 motorcycles out of 40 and 34 out of 80 are the same share.
From tally to AI
Frequency tables are everywhere in AI. Before training a model, data scientists count how many examples there are of each label, because an imbalance, such as 950 photos of healthy leaves and 50 of diseased ones, warns that the model may struggle with the rare class. Language models are built on counts of which words follow which. Counting carefully is where understanding data begins.
Frequency tables are also the first step towards charts. Each row of the table becomes one bar of a bar graph, and the frequency becomes the height of the bar. When values are numbers spread over a wide range, such as heights or marks, they are first grouped into ranges, like 0 to 10 and 11 to 20, and the frequency of each range is counted. That grouped table is what a histogram shows.
How it works
- Fix the categories in advance.
- Tally as you observe; do not rely on memory.
- Cross every fifth stroke.
- Check that the total equals the number of observations.
Try it yourself
Count the answers and build a table
Tally each answer, watch the frequency table and bar chart grow, then read the result.
Traffic count
Do this from a safe place, such as a balcony or a school gate, with an adult.
- Draw a table with rows for motorcycle, car, auto-rickshaw, bus, cycle and other.
- For ten minutes, make a tally mark for each vehicle that passes.
- Count the tallies to find each frequency and the total.
- Work out which vehicle was most common and what share of the total it made.
Think about it
- Would the counts be different at another time of day? Why?
- What question could this data help answer?
Common mistakes
- Forgetting an "other" category, so some observations have nowhere to go.
- Counting from memory after the event instead of tallying as you go.
- Not checking that the frequencies add up to the total.
Check yourself
Summary
- Tally marks count quickly in groups of five.
- A frequency table lists each category with its count and a total.
- Counting carefully is the first step of data analysis, including for AI.
Finished reading?
Sources: DataBasic.io: tools and activities for data literacy · Ministry of Statistics and Programme Implementation (India)