Before you begin, five plain-word reminders so no symbol ambushes you later. Every piece of notation this page uses is defined right here first.
Recall What "frequency" means, and the symbol
f
Frequency and f ::: Frequency is the count — how many times a value or category shows up. We write it with the letter f. When there are several categories we tag each one with a little number: f1 is the first category's count, f2 the second, and in general fi means "the count of the i-th category" (read i as "which row we're looking at").
Recall What the symbol
∑f (sigma notation) means
∑f ::: The tall Greek letter ∑ (capital "sigma") is a shorthand for "add them all up". So ∑f=f1+f2+f3+… means "take every category's frequency and total them". It is just an economical way to say "the grand total of all the counts" without writing out a long +⋯+ chain.
Recall What "sample size" and the symbol
n mean
Sample size n ::: The number of observations you actually made — how many people you surveyed, dice you rolled, plants you measured. We call it n. Because each observation adds exactly 1 to exactly one category, the grand total of counts equals the sample size: ∑f=n.
Recall What a tally is, and how a crossed group of five looks
Tally ::: A single upright mark (∣) that stands for one observation, drawn the instant it happens. Four marks stand side by side (∣∣∣∣); the fifth mark is drawn diagonally across those four, turning them into a bundle worth 5. In this note we picture that crossed bundle as "∣∣∣∣" and refer to it in words as a crossed group of five — that diagonal stroke is the whole convention (see the figure below).
Recall What cumulative frequency means
Cumulative frequency ::: A running total of frequencies read from the top of the table downward. The cumulative frequency at row i is Ci=f1+f2+⋯+fi=∑k=1ifk (add every count from the first row down to row i). Because you only ever add non-negative counts, this running total can never shrink as you move down.
Here is what those tally bundles and their frequency table look like — refer back to the crossed diagonal whenever a question below mentions a "crossed group of five":
And here is the same frequency table drawn as a bar chart, so "frequency" is something you can see as height, not just a number:
A survey I run myself is primary data even if I copy the questions from a magazine.
True — "primary" is about who collects the responses, not who wrote the questions. You gathered the answers firsthand, so the data is yours.
Government census figures become primary data the moment I download and analyse them.
False — downloading does not change who collected them. The census was collected by someone else, so for you it stays secondary no matter how you use it.
The total of all frequencies must equal the number of observations you made.
True — every observation adds exactly one to exactly one category, so ∑f=n (the sample size). If it doesn't, you miscounted.
Two categories with the same frequency mean the survey was done wrong.
False — ties are perfectly normal. Equal counts just say those categories are equally common; nothing about the method is broken.
A relative frequency can be greater than 1.
False — it is a part divided by the whole, fi/∑f, so it always lies between 0 and 1 (or 0% to 100%).
Cumulative frequency can go down as you move down the table.
False — each Ci=∑k=1ifk is a running total of non-negative counts, so it can only stay the same or increase, never fall.
Secondary data is always cheaper and better than primary data.
False — it is usually cheaper, but it may not match your exact question, may be outdated, or may hide how it was collected. Cheaper does not mean better-fitting.
The last cumulative frequency in a table always equals ∑f.
True — by the last row you've added every frequency, so the running total Clast=∑f=n.
A student writes a tally group as ∣∣∣∣∣ (five upright marks, none crossed) and calls it "5".
The count is right but the convention is broken — the fifth mark must be a diagonal stroke across the first four (a crossed group of five, as in the top figure). Uncrossed groups of five are exactly what your eyes misread once the marks pile up.
A frequency table lists Apple =11, Banana =8, Orange =4, and a "Total" row of 22.
The total is wrong: ∑f=11+8+4=23, not 22. A total that doesn't match the summed frequencies is a red flag that a count (or the addition) slipped.
Someone says "Apple is 47.8%, so about half of all people prefer apples."
The percentage is 11/23within this class only. It describes this sample, not all people. Generalising a small classroom survey to everyone is an overreach.
A tally chart has a "Tally" column but no "Frequency" column, and the student says it's finished.
A tally chart is the live-recording stage; it isn't analysis-ready until you count the marks into numbers. The missing frequency column is the whole point of the next step.
A relative-frequency column reads 10%,15%,35%,30%,10% and totals 110%.
Impossible — proportions of one whole must sum to 100%. One percentage is mistyped; recompute each fi/∑f.
A student records "temperature from a weather website" and labels it primary because they looked it up.
Looking something up is not collecting it. The website measured the temperature, so it is secondary data for the student.
Why do we cross out every fifth tally instead of just drawing marks in a row?
Because the human eye can instantly recognise a crossed group as "5" but cannot reliably count a long line of loose marks. Grouping in fives turns counting into simple 5×(groups)+leftovers.
Why convert a tally chart into a frequency table at all?
Numbers are far quicker to compare, add, and turn into percentages than piles of marks. The frequency table is the clean form that later tools (bar charts, the mean) actually feed on.
Why do we bother with relative frequency when we already have the counts?
Counts from different-sized samples aren't comparable — "11 out of 23" versus "15 out of 50" is hard to judge. Turning them into 47.8% vs 30% makes the comparison instant.
Why is cumulative frequency useful if the plain frequencies are already there?
It answers "how many up to and including this category?" in one glance — questions like "how many students scored ≤60?" — without re-adding rows each time.
Why might a scientist still collect primary data when good secondary data exists?
Because secondary data may not answer their exact question, may be out of date, or may have unknown collection quality. Primary data is tailored and its origin is fully controlled.
Why must you always check that ∑f=n?
It is a cheap self-test: since each observation counts once, any mismatch between the total of counts and the sample size means a lost, double-counted, or miscategorised value — caught before it corrupts the whole analysis.
If a category gets zero observations, do you still list it in the frequency table?
Yes if it's a genuine option you were tracking — a frequency of fi=0 is real information ("nobody chose this"). Omitting it can mislead readers into thinking the option didn't exist.
What is the tally for a category with exactly five observations?
One crossed group of five (four uprights with a diagonal fifth) — a full bundle with no leftover single marks. Frequency =1×5+0=5.
You survey 0 people. What does the frequency table look like?
Every frequency is 0 and ∑f=0. Relative frequencies are undefined — you'd be dividing by ∑f=0, which has no meaning.
One respondent picks two favourite fruits. How does that break ∑f=n?
It adds two tally marks for one person, so ∑f exceeds n, the number of people. Fix the design first: force one choice, or count "responses" (not "people") as your total.
A category has 23 observations — how many complete groups of five and how many leftovers?
23=4×5+3: four crossed groups plus three single marks. This is exactly the "(groups)×5+remaining" rule at work.
Is the "most frequent category" always a meaningful mode if two categories tie for the top?
No single mode exists then — the data is bimodal (two modes). Reporting just one would hide that two categories are equally the most common.