Cumulative Frequency for Sengkang Families: Read Running Totals and Explain Estimates

Imagine five fictional reading sessions lasting 8, 12, 12, 19 and 23 minutes. A frequency table can show how many sessions have each duration. A cumulative table answers a different question: how many sessions lasted at most a given time? At twelve minutes, three sessions have been counted because eight and both twelves meet that boundary. The number is not simply the frequency of twelve.

The word cumulative means the count builds as the boundary increases. Once an observation belongs to the count, it remains included at every higher boundary. This explains why a correctly ordered cumulative frequency cannot decrease. The individual frequencies can rise or fall, but the running total either increases or stays the same.

Begin with paper cards representing the sessions. Move a boundary from ten minutes to fifteen and then twenty. Add the newly qualifying cards without removing earlier ones. The physical action makes the distinction visible before the learner needs a formula or a graph. It also helps with interpreting a count that stays unchanged when no new observation enters.

Ask the child to complete a sentence: “At this boundary, the cumulative frequency tells me how many observations are ___.” Include the direction and units. “Three” alone is not yet an interpretation. “Three sessions lasted at most twelve minutes” connects the number with the intended set and prepares the learner to read more complicated tables accurately.

Contents · Next chapter

One row in a dataset may represent a person, a session, an object or a measured event. A cumulative frequency counts those observations. In the reading example, five sessions do not necessarily mean five different readers. One person may contribute several sessions. The chart’s interpretation depends on what was recorded, not on a guess from the topic.

For a fictional set of ten packets, each observation might be the mass of one packet in grams. A cumulative count of seven at fifty grams then means seven measured packets weigh at most fifty grams. It does not mean that seven grams were measured, nor does it prove that every packet made by the same process will behave similarly.

Write the variable and unit beside the table before calculating. “Mass, grams” describes the horizontal measurement; “number of packets” describes frequency. Keep those quantities distinct when reading a graph. A learner who says the median is twenty packets may have read the vertical count without returning to the horizontal measurement scale.

For family practice, give the same numerical table two possible stories and ask what changes in the interpretation. The running totals can be identical while their units and conclusions differ. This helps a child see that mathematical processing and contextual explanation are separate but linked tasks. A correct number should always have an accurate answer to “number of what?”

Contents · Previous chapter · Next chapter

Cumulative frequency requires a meaningful order. Durations can be arranged from shortest to longest, and scores from lowest to highest. A running total of favourite colours in an arbitrary colour order usually does not answer a meaningful threshold question. The table must be organised around a variable whose earlier and later values have a sensible relationship.

For discrete scores 1, 2, 3 and 4 with frequencies 2, 5, 1 and 4, the cumulative counts are 2, 7, 8 and 12. If the rows were rearranged as 4, 1, 3 and 2, simply adding down would still produce a running total, but its intermediate rows would no longer mean scores at or below the listed boundary. The ordering is part of the definition.

Some ordered categories, such as low, medium and high, can have meaningful cumulative counts if their ranking is explicit. But not every label that can be alphabetised is an ordered measurement. Ask what “at or below this category” would mean before deciding whether a cumulative representation serves the question.

For practice, sort several fictional variables into suitable and unsuitable candidates. Include score, duration, colour and a clearly defined ordered rating. Explain the decision before building a table. This prevents a learner from applying a procedural rule to any column of numbers or labels without first checking whether the resulting boundary statements would actually make sense.

Contents · Previous chapter · Next chapter

Suppose the frequencies for scores 1, 2, 3 and 4 are 2, 5, 1 and 4. The frequency of score two is five, while the cumulative frequency at score two is seven. The first counts one score category. The second includes score one and score two. These two answers can both be correct because the questions differ.

Ask the learner to translate “exactly two” and “at most two” into sets before reading the numbers. Exactly two selects one row. At most two includes every row up to that value. If the question says “less than two,” score two itself is excluded, leaving only the lower category. Boundary words change the requested count.

A useful comparison is to write frequency and cumulative frequency side by side. For each row, explain what is newly added and what has already been counted. Do not assume the two columns should have similar shapes. A small ordinary frequency late in the table can accompany a large cumulative total because many earlier observations are already included.

For a family check, ask which column should be used to find the most common single score. The answer is ordinary frequency, not the largest cumulative total. The cumulative maximum normally occurs at the final boundary regardless of which category is most common. This distinction helps prevent a familiar-looking number from answering a different question than the one actually asked.

Contents · Previous chapter · Next chapter

Take four fictional duration groups with frequencies 5, 11, 14 and 10. The running totals are 5, 16, 30 and 40. Start with the first count. Then add the second frequency to the first cumulative count, and continue. At every step add only the new group’s frequency, because the previous running total already includes all earlier groups.

An error would be to add all previous cumulative totals together. At the third row, five and sixteen are not independent groups; sixteen already includes five. Adding both would count the first group twice. The difference between a group count and a running count matters as soon as more than two rows are involved.

Check the last cumulative frequency against the total number of observations. For the example, five plus eleven plus fourteen plus ten is forty. The final cumulative count should also be forty if the displayed groups include the complete dataset. A mismatch suggests a copying error, missing group or disagreement about which observations belong in the table.

For family practice, leave one cumulative entry blank and ask the learner to reconstruct it from the previous total and the current frequency. Then ask for a verbal interpretation using a group boundary. Calculation alone is useful, but the learner should also know what the resulting number counts and why earlier observations remain part of it.

Contents · Previous chapter · Next chapter

If the cumulative frequencies are 5, 16, 30 and 40, the second group contains 16 − 5 = 11 observations. The third contains 30 − 16 = 14, and the fourth contains 40 − 30 = 10. The first contains five if the running count starts at zero before the first displayed group. Adjacent differences recover the new observations added at each step.

This reverses the construction process. Addition builds a running total from group counts; subtraction recovers the increment between two cumulative boundaries. A learner who reports thirty for the third group has read the whole count up to its end instead of the number that entered within that particular interval.

If consecutive cumulative values are equal, their difference is zero. The group has no recorded observations. That is valid and can produce a flat graph segment. A negative difference is not a valid ordinary group frequency in an increasing-boundary count, so it signals a data or reading problem that needs investigation.

For practice, provide only a cumulative column and ask the child to recover the frequencies and then rebuild the column. This round trip checks both operations and provides a useful audit. If the reconstructed frequencies fail to add to the final cumulative count, revisit the first row convention or the subtraction order rather than adjusting a number arbitrarily.

Contents · Previous chapter · Next chapter

Consider positive durations grouped as 0 < t ≤ 10, 10 < t ≤ 20, 20 < t ≤ 30 and 30 < t ≤ 40 minutes. Exactly ten belongs in the first group, exactly twenty in the second and exactly thirty in the third. The inequalities prevent overlap at shared boundaries. Read them before placing or counting a boundary value.

Other tables use different endpoint conventions. A group labelled 10 ≤ t < 20 includes ten but excludes twenty. A table of rounded measurements may use class boundaries derived from the rounding rule. Do not silently switch conventions because one familiar format appeared in an earlier worksheet. The stated interval defines which observations are included.

In the positive-duration example, zero is excluded deliberately. If a real task permits zero-minute observations, the first interval must account for them. Check the domain and the grouping together. A neatly accumulated table can still lose observations if its intervals leave gaps or assign an endpoint to two groups.

For a family contrast, choose three values just below, exactly at and just above twenty minutes. Place each in the correct interval under two different conventions. Explain why only the boundary value changes group. This small exercise often repairs confusion more effectively than a lengthy calculation with no exact endpoint cases. The arithmetic depends on a valid classification first.

Contents · Previous chapter · Next chapter

Using the example totals, sixteen observations lasted at most twenty minutes, and forty were recorded altogether. Therefore twenty-four lasted more than twenty minutes. Subtracting from the total gives the complementary upper group. The words more than exclude exactly twenty, which is already included in at most twenty under the stated convention.

The number with 20 < t ≤ 30 is thirty minus sixteen, or fourteen. The first cumulative count includes values at most thirty; the second removes every value at most twenty. The result has an open lower boundary and a closed upper boundary. Saying only “between twenty and thirty” may obscure those exact inclusion choices.

The number with t ≥ 20 is not necessarily twenty-four. That would also include observations exactly equal to twenty, whose count is not separately known from the grouped table. A learner who assumes a group boundary has no observations may produce a confident answer unsupported by the data. Exact inclusion can matter even when the numerical threshold is identical.

For practice, rewrite a threshold question as inequalities before calculating. Then identify which two sets are being compared and why subtraction is valid. This habit is particularly useful for upper-tail questions, where at least and more than can differ. Read the evidence available rather than invent a count at an exact boundary merely to complete a calculation.

Contents · Previous chapter · Next chapter

For the example groups ending at ten, twenty, thirty and forty, plot points (10, 5), (20, 16), (30, 30) and (40, 40). The horizontal coordinate is the upper boundary, and the vertical coordinate is the count up to that boundary. A starting point (0, 0) is appropriate for this complete positive-duration dataset because no observation is at or below zero.

Do not plot group midpoints as the horizontal positions in this cumulative construction. A point at fifteen with count sixteen would suggest that sixteen observations are at most fifteen minutes, which is not what the table establishes. The count belongs to the end of the interval because it includes the full group up to that boundary.

Different tasks may specify joining points with straight segments or drawing a smooth cumulative curve. Follow the instruction and recognise that an intermediate read is generally an estimate for grouped data. The plotting convention does not reveal where each original observation occurred inside its interval. It gives a readable model of the accumulated count.

For a family activity, plot the original points on labelled axes and read each exact plotted boundary back into words. Then choose a value between two boundaries and ask whether the count is exact or estimated. This checks both construction and interpretation, preventing the learner from treating every location on a drawn line as an original measured fact.

Contents · Previous chapter · Next chapter

The horizontal axis must represent the measured variable with its units. The vertical axis represents cumulative count, or cumulative percentage if the task explicitly uses that version. Choose consistent increments along each axis. The two axes need not have the same numerical increment because they represent different quantities. Both scales should make boundaries and totals easy to read.

For the forty-observation example, a vertical scale from zero to forty in steps of five is convenient. A horizontal scale from zero to forty minutes in steps of ten makes the class boundaries clear. Smaller grid divisions can support intermediate estimates. Labelling a scale clearly matters more than filling every available square on the page.

Do not compare the visual steepness of two graphs without checking their scales. A stretched vertical axis can make a curve look steeper even when the data are unchanged. If the purpose is comparison, use consistent axes or compare specific numerical counts and percentages rather than making claims from appearance alone.

For a family check, ask the learner to read one minor grid interval before plotting or estimating. If a large division is five observations and contains five equal smaller divisions, each small division represents one. A scale-reading mistake can affect every later interpretation, so identify it before assuming the child misunderstood the statistics themselves.

Contents · Previous chapter · Next chapter

If consecutive cumulative points have the same vertical value, no observations were added between those stated boundaries. For example, counts twelve at twenty minutes and twelve at thirty minutes imply zero observations in 20 < t ≤ 30 under the example convention. The flat segment is information, not a graphing failure.

An increasing portion means new observations enter as the boundary increases. With equal horizontal interval widths, a larger vertical increase means a larger frequency in that interval. With unequal widths, compare actual differences and boundaries rather than assuming that one visual steepness comparison directly gives group counts. The shape can suggest concentration, but the exact calculation uses the stated axes and intervals.

A downward cumulative section is inconsistent with this increasing-boundary definition if the same dataset and count are being used. Check whether the graph actually shows a different measure, such as an upper-tail count, or whether a value was plotted incorrectly. Do not diagnose an error solely because a graph called cumulative uses a different explicit orientation.

For family practice, sketch two consecutive equal totals and explain the interval count. Then invent one impossible decreasing pair and identify the contradiction. This helps the learner distinguish valid zero frequency from invalid negative frequency and understand why the direction of accumulation places a mathematical constraint on the graph.

Contents · Previous chapter · Next chapter

To estimate how many observations are at most twenty-five minutes, start at twenty-five on the horizontal axis. Move up to the cumulative graph and then across to the vertical count scale. If the graph joins (20, 16) and (30, 30) with a straight line, the midpoint estimate is twenty-three observations. This is an interpolation, not a known exact count from the grouped table.

The exact grouped facts tell us that sixteen observations are at most twenty and thirty are at most thirty. They do not tell us how the fourteen observations inside that group are distributed around twenty-five. A straight line assumes a constant rate of accumulation within the interval for the purpose of an estimate. A different within-group distribution could give a different exact count.

If the requested threshold is one of the stated class boundaries, use the exact table value when available. Reading a graph by eye may introduce unnecessary rounding. At twenty minutes, sixteen is exact for this table. At twenty-five, the graph-based estimate should be labelled appropriately. Do not treat a smooth-looking line as proof of exact individual values.

For practice, write the two directions in words: threshold to count starts on the horizontal axis; count to threshold starts on the vertical axis. This distinction prepares the learner for quartiles and prevents a frequent error of reading the requested measurement directly from the wrong scale.

Contents · Previous chapter · Next chapter

To estimate a duration at a stated cumulative position, start on the vertical axis. For example, at cumulative count twenty in the forty-observation example, draw across to the graph and down to the horizontal axis. The resulting duration is the model-based boundary at which the graph reaches that cumulative level. It is measured in minutes, not observations.

With straight-line interpolation between (20, 16) and (30, 30), count twenty is four above sixteen out of a total increase of fourteen. Move four-fourteenths of the ten-minute interval above twenty. The estimate is about 22.86 minutes. A hand-drawn graph may support only a rounded reading such as about twenty-three minutes, depending on its scale.

The calculation assumes linear accumulation within that group. It does not reconstruct the exact twentieth observation or prove the true sample median equals 22.86. Grouped data lost the individual values. State what the question asks: an estimate from the supplied graph or interpolation method, rather than an exact order statistic from raw data.

For family practice, mark the chosen vertical position and explain why it lies between two cumulative totals. Then locate the corresponding horizontal interval before calculating. This range check catches answers read from the wrong class or wrong axis. An estimate should not fall outside the interval whose cumulative counts bracket the requested position.

Contents · Previous chapter · Next chapter

For a conventional grouped cumulative-frequency graph with forty observations, the median estimate is read at cumulative level twenty. The vertical level is half the total; the answer is the horizontal value reached at that level. In the example’s straight-line model, that value is about 22.86 minutes. Twenty is the count position, not the duration.

For raw ordered data, the exact median follows the convention for the middle observation or middle pair. With forty raw observations, the usual exact median averages the twentieth and twenty-first values. A grouped graph does not give those exact values. Reading at half the total is a common grouped estimation method, not a claim that the raw-data procedure and grouped model are identical.

Follow the convention specified in the task. Different courses and software can use different sample quantile definitions, particularly for small raw datasets. The main family habit is to identify the data form and the instructed method before calculating. Do not transfer a raw-list position rule mechanically to every grouped graph question.

Ask the learner to state three separate items: total observations, chosen cumulative level and estimated measurement. For the worked example they are forty, twenty and about 22.86 minutes. This simple separation helps identify whether a wrong answer came from finding the position, reading the graph or labelling the resulting value.

Contents · Previous chapter · Next chapter

For the same grouped graph, the lower quartile is commonly estimated at one-quarter of the total and the upper quartile at three-quarters. With forty observations, the corresponding vertical levels are ten and thirty. Read across from those counts and down to the measurement scale. The quartile values are durations in this example, not the counts ten and thirty themselves.

Count ten lies between five at ten minutes and sixteen at twenty. Under straight-line interpolation, it is five-elevenths of the way through that ten-minute interval, giving about 14.55 minutes. Count thirty is reached at the boundary thirty minutes in the supplied graph. The upper-quartile estimate is therefore thirty minutes under that model and convention.

The fact that a graph estimate lands on a known class boundary does not guarantee the exact raw-data quartile is that number. The table establishes how many observations lie up to the boundary, not their exact internal positions. Keep the label estimate when the question uses grouped data or a plotted model.

For practice, ask the learner to locate the relevant cumulative level before reading any horizontal value. Check that the lower quartile does not exceed the median and that the median does not exceed the upper quartile. This ordering is a useful sanity check, though it does not by itself verify that the correct positions or scales were used.

Contents · Previous chapter · Next chapter

The interquartile range is upper quartile minus lower quartile. Using the worked grouped estimates, it is thirty minus about 14.55, or about 15.45 minutes. It measures the width between the two quartile values. The answer has the same unit as the measured variable, rather than a count or percentage unit.

Do not subtract the cumulative levels thirty and ten and report twenty minutes. That subtracts two vertical counts, not two measured durations. The calculation needs the horizontal quartile estimates first. This is another example of why the position and value must remain separate in working and explanation.

The interquartile range describes central spread, while the full range concerns extremes. A dataset can have a large full range because of an unusual extreme yet a modest central spread. The grouped table may not provide exact minimum and maximum observations, so avoid treating the outer class boundaries as verified extremes unless the task explicitly gives that interpretation.

For family practice, compare two imaginary sets with the same median but different interquartile ranges. Ask what that suggests about their central spreads and what it does not establish about every individual observation. The learner should be able to explain the measure in words rather than treat it as merely another subtraction with no relation to the data distribution.

Contents · Previous chapter · Next chapter

Suppose fictional group A has sixteen of forty observations at most twenty minutes. Group B has twelve of twenty at that boundary. A has the larger count, but B has the larger proportion: sixty percent compared with forty percent. Counts and percentages answer different comparison questions, especially when the totals differ.

To form a cumulative percentage, divide each running count by the relevant total and multiply by one hundred. In the original example, the percentages are 12.5, 40, 75 and 100. The last is one hundred percent if the groups cover the entire dataset. Rounding earlier percentages can create small differences, so use the original counts when exact comparison is needed.

Keep denominators visible. If a learner divides group B’s twelve by group A’s forty, the resulting percentage describes neither group properly. A shared threshold does not create a shared denominator. Compare the proportion within each group, and state how the observations were collected if drawing conclusions beyond the fictional data.

For a family task, ask which group had more qualifying observations and which had a higher qualifying share. Require two separate sentences. This makes the distinction concrete and prevents a student from interpreting “higher” without naming what is higher. A truthful comparison may need both the count and the proportion rather than one number presented as the whole story.

Do not infer a complete distribution from one threshold comparison. Group B having a higher proportion at most twenty minutes does not establish that B is faster at every other threshold or that its interquartile range is smaller. Those conclusions require other parts of the tables or graphs. Two cumulative percentage curves can cross when one group has more observations below an early boundary but fewer below a later one. A fair summary should name the threshold actually compared and avoid extending that local result to every part of the distribution. To compare medians, read each group’s fifty-percent level separately and state any estimation assumption.

Contents · Previous chapter · Next chapter

Inspect four fictional errors. A reads a cumulative count of thirty as the third group’s frequency. B plots a cumulative count at the class midpoint. C reports the vertical half-total as the median duration. D treats a value read inside a grouped interval as an exact original measurement. Each error concerns a different step.

A needs adjacent subtraction: thirty minus sixteen equals fourteen. B needs the appropriate upper boundary for the stated less-than-or-equal cumulative graph. C must read across and down to the measurement axis after finding the half-total. D must distinguish grouped evidence from the interpolation model. A correct arithmetic step elsewhere would not repair any of these decisions automatically.

Ask the learner to name the information available before choosing a repair. Is the original data list present, or only groups? Does the threshold match a boundary? Is the question asking for a count, a proportion or a measurement? These checks make the task clearer than an instruction to read the graph more carefully without specifying what to look for.

For a fresh contrast, change only one feature: move a threshold inside a group, swap a count question for a measurement question or change the dataset total. Then ask what changes in the method. A small targeted example can show whether the learner has understood the decision rather than simply copied the correction.

Contents · Previous chapter · Next chapter

Contents · Previous chapter · Next chapter

Contents · Previous chapter · Next chapter

Contents · Previous chapter · Next chapter

Contents · Previous chapter · Next chapter

Start with one table and ask four questions: what is being counted, what boundary is specified, does equality belong, and is the answer a count or a measurement? Then choose addition, subtraction, a graph read or a percentage calculation. This routine makes method choice follow the actual question rather than an automatic response to seeing a cumulative graph.

Change one feature at a time in later practice. Replace at most with more than, change the group total or move a threshold from an exact boundary into an interval. Ask what changes in the reasoning. Controlled contrasts reveal misunderstandings about inclusion, denominators or estimation without adding many new calculations at once.

Keep a short mistake note with the first unstable decision: used a group count instead of a running count, plotted a midpoint, read the wrong axis or claimed exactness from grouped data. A fresh task should test that same decision. Copying the corrected answer is weaker evidence than explaining a new case with different numbers.

When seeking help, bring the table, graph and the learner’s interpretation. A teacher or tutor can then separate arithmetic, scale reading, boundary language and evidence limits. The goal is a learner who can justify what the graph tells us and explain what it cannot tell us, rather than merely draw an attractive curve or produce a plausible-looking number.

Contents · Previous chapter · Next chapter

Does cumulative frequency ever go down? A correctly constructed increasing-boundary count for one fixed dataset does not decrease. It can stay flat when no observations enter between boundaries. If a display falls, inspect whether it shows a different quantity, such as an upper-tail count, or whether a value was copied or plotted incorrectly.

Is every graph reading exact? No. A stated boundary count can be exact for the grouped table, while a reading inside a group is usually an estimate based on the graph or interpolation method. Grouping removes individual detail. Do not report an exact raw-data value merely because the graph can be read to several decimal places.

Why are quartiles read horizontally after finding a vertical level? The vertical coordinate counts observations. The quartile is a value of the measured variable, so the final answer comes from the horizontal axis. The chosen vertical level tells the reader where to begin the graph read, not what unit the final measurement should have.

What comes next? Revisit ordinary frequency if running totals are unclear, coordinates if graph reading is unstable and percentages if datasets have different sizes. The linked formal statistics guide offers broader progression into box plots and related measures. Keep the boundary, count, measurement unit and estimate label visible. Those habits make cumulative frequency a clear way to communicate evidence rather than a set of graph-reading moves performed without context.

References and further reading

The worked scenes and practice sets are original teaching examples. These references offer formal definitions or documented usage; the connected eduKate guides provide foundations and further study.

OpenStax: frequency and frequency tables · Secondary 3 guide to cumulative frequency and box plots · Scatter plots for families

Contents · Previous chapter · Continue at the learning hub