Free Mathematics Class 10 CBSE notes · practise this chapter with an AI quiz

← All study notes

Three Different Methods Must Give the Same Average, and One of Them Is Far Quicker

Find the mean of grouped data by the direct method using class marks, shift the numbers with an assumed mean, divide them down with the step-deviation method, decide which method the data suits, and work backwards to a missing frequency.

How do you average data when you no longer know the individual numbers?

Suppose a survey records the ages of 3030 people, but instead of the 3030 ages it reports only how many fall in each band: two people in the band 1010–2525, three in 2525–4040, seven in 4040–5555, and so on. The individual ages are gone. You cannot add them up, because you do not know them.

So you do the next best thing: assume everyone in a band sits at its middle. The two people in 1010–2525 are treated as being 17.517.5 each; the seven in 4040–5555 as 47.547.5 each. That middle value is the class mark:

class mark=lower limit+upper limit2\text{class mark} = \frac{\text{lower limit} + \text{upper limit}}{2}

Then the ordinary averaging method works again, because you have a single number for each group and a count of how many share it.

The answer is an estimate, and it is worth being honest about that. Some people in a band are above its middle and some below; the assumption is that those cancel out across the band. For a large data set they very nearly do, which is why this is the standard method and not a compromise.

From there the chapter gives you three ways to do the arithmetic, and all three must give the same answer:

- The direct method — multiply each class mark by its frequency and divide
- The assumed mean method — shift all the numbers down by a convenient amount first
- The step-deviation method — shift them and then divide them down by the class width

They are not three different means. They are three different routes to one number, and the reason for having three is purely arithmetic: the third one replaces multiplications like 7×47.57 \times 47.5 with multiplications like 7×07 \times 0. On a distribution with equal class widths it is dramatically faster, and that is why it exists.

This page covers the first part of the CBSE Class 10 Maths chapter on statistics: the mean of grouped data by all three methods, choosing between them, and finding a missing frequency.
Formula

What are the three formulas for the mean of grouped data?

All three compute the same mean; they differ only in what they multiply the frequencies by.

The direct method, using the class marks xix_i:

xˉ=∑fixi∑fi\boxed{\bar{x} = \frac{\sum f_i x_i}{\sum f_i}}

The assumed mean method, where aa is any convenient value — usually a class mark near the middle — and di=xi−ad_i = x_i - a:

xˉ=a+∑fidi∑fi\boxed{\bar{x} = a + \frac{\sum f_i d_i}{\sum f_i}}

The step-deviation method, where hh is the common class width and ui=xi−ahu_i = \frac{x_i - a}{h}:

xˉ=a+h(∑fiui∑fi)\boxed{\bar{x} = a + h\left(\frac{\sum f_i u_i}{\sum f_i}\right)}

Why the second one works. Subtracting the same aa from every value lowers the mean by exactly aa, so adding aa back at the end restores it. The mean of the shifted data plus the shift equals the mean of the original data — which is a statement about averages that holds for any data at all.

Why the third one works. Dividing every deviation by hh divides the mean of the deviations by hh, so multiplying by hh at the end restores it. The step-deviation method is the assumed mean method with a second, undone, simplification.

Notice the structural similarity. Each formula is "something ++ a correction", and the correction is a weighted average of the deviations. In the direct method the something is 00 and the deviations are the values themselves.

The condition on the third method, which examiners ask about. The step-deviation method needs all class widths to be equal
, because a single hh has to divide every deviation. If the classes have unequal widths, use the direct or the assumed mean method — and saying so is itself an examinable point.

A convention that must be stated for the first method to be well defined. Class marks are computed from the class limits as given. If the classes are written as 1010–2525, 2525–4040 and so on, they are already continuous and the marks are 17.517.5, 32.532.5 and so on. If they are written as 1111–1313, 1414–1616, with gaps, they must first be made continuous by taking the class boundaries halfway between the limits — and a question that gives gapped classes is testing exactly that step.

How do you find the mean by the direct method?

Write the class mark for every class, multiply each by its frequency, add both columns, and divide.

Worked example — the distribution used throughout this page. Find the mean of the following distribution of 3030 observations.

- 1010–2525: frequency 22, class mark 17.517.5, product 2×17.5=352 \times 17.5 = 35
- 2525–4040: frequency 33, class mark 32.532.5, product 3×32.5=97.53 \times 32.5 = 97.5
- 4040–5555: frequency 77, class mark 47.547.5, product 7×47.5=332.57 \times 47.5 = 332.5
- 5555–7070: frequency 66, class mark 62.562.5, product 6×62.5=3756 \times 62.5 = 375
- 7070–8585: frequency 66, class mark 77.577.5, product 6×77.5=4656 \times 77.5 = 465
- 8585–100100: frequency 66, class mark 92.592.5, product 6×92.5=5556 \times 92.5 = 555

The two totals:

∑fi=2+3+7+6+6+6=30\sum f_i = 2 + 3 + 7 + 6 + 6 + 6 = 30
∑fixi=35+97.5+332.5+375+465+555=1860\sum f_i x_i = 35 + 97.5 + 332.5 + 375 + 465 + 555 = 1860

The mean:

xˉ=186030=62\bar{x} = \frac{1860}{30} = 62

Check the answer for plausibility. The class marks run from 17.517.5 to 92.592.5, so the mean must lie somewhere between them — and 6262 does. More usefully, the frequencies are heavier in the upper classes, so the mean should sit above the middle of the range, which is 5555. It does, at 6262. That two-second check catches a misplaced decimal point or a column added wrongly.

How the class marks were found.
Each is the average of the two limits:

10+252=17.5,25+402=32.5,40+552=47.5\frac{10 + 25}{2} = 17.5, \qquad \frac{25 + 40}{2} = 32.5, \qquad \frac{40 + 55}{2} = 47.5

And they rise by 1515 each time, which is the class width. That regular step is what makes the step-deviation method possible, and noticing it early tells you which method to reach for.

Where the direct method costs you.
The products 332.5332.5, 375375, 465465 and 555555 all involve multiplying a two-digit frequency by a number with a decimal. On a distribution with larger class marks — thousands of rupees, say — this becomes genuinely slow and error-prone, and that is the practical reason the other two methods exist.

When the direct method is nevertheless the right choice. Use it when:

- The class marks are small whole numbers, so the products are easy
- The class widths are unequal, which rules the step-deviation method out
- There are only two or three classes, so no method saves much

One boundary case about class marks. A class written as 1010–2525 and the next as 2525–4040 appear to share the value 2525. The convention is that an observation of exactly 2525 is counted in the class 2525–4040, so each class includes its lower limit and excludes its upper. The totals are unaffected either way, but the convention has to be stated when a question asks how a boundary value is treated.

How do the assumed mean and step-deviation methods speed this up?

Subtract a convenient value from every class mark, and then divide by the class width. Both operations are undone at the end, so the mean is unchanged.

Worked example 1 — the assumed mean method on the same data. Take a=47.5a = 47.5, the class mark of the middle-ish class, and compute di=xi−47.5d_i = x_i - 47.5.

- 1010–2525: f=2f = 2, x=17.5x = 17.5, d=−30d = -30, product 2×(−30)=−602 \times (-30) = -60
- 2525–4040: f=3f = 3, x=32.5x = 32.5, d=−15d = -15, product −45-45
- 4040–5555: f=7f = 7, x=47.5x = 47.5, d=0d = 0, product 00
- 5555–7070: f=6f = 6, x=62.5x = 62.5, d=15d = 15, product 9090
- 7070–8585: f=6f = 6, x=77.5x = 77.5, d=30d = 30, product 180180
- 8585–100100: f=6f = 6, x=92.5x = 92.5, d=45d = 45, product 270270

∑fidi=−60−45+0+90+180+270=435\sum f_i d_i = -60 - 45 + 0 + 90 + 180 + 270 = 435
xˉ=47.5+43530=47.5+14.5=62\bar{x} = 47.5 + \frac{435}{30} = 47.5 + 14.5 = 62

The same mean of 6262, reached with much smaller numbers. And notice what happened to the class containing aa: its deviation is zero, so it contributes nothing at all. Choosing aa to be the class mark of the class with the largest frequency kills the biggest product outright, which is the whole reason for choosing aa from the middle of the data rather than from an end.

Worked example 2 — the step-deviation method on the same data. Every deviation above is a multiple of 1515, which is the class width. Divide by it: ui=di15u_i = \frac{d_i}{15}.

- 1010–2525: f=2f = 2, u=−2u = -2, product −4-4
- 2525–4040: f=3f = 3, u=−1u = -1, product −3-3
- 4040–5555: f=7f = 7, u=0u = 0, product 00
- 5555–7070: f=6f = 6, u=1u = 1, product 66
- 7070–8585: f=6f = 6, u=2u = 2, product 1212
- 8585–100100: f=6f = 6, u=3u = 3, product 1818

∑fiui=−4−3+0+6+12+18=29\sum f_i u_i = -4 - 3 + 0 + 6 + 12 + 18 = 29
xˉ=47.5+15×2930=47.5+14.5=62\bar{x} = 47.5 + 15 \times \frac{29}{30} = 47.5 + 14.5 = 62

The same 6262 again, and this time every number in the working column was a single digit. Compare the three sums you had to compute: 18601860, then 435435, then 2929. That is the whole case for the step-deviation method.

Which method is most efficient for this data, and why. The step-deviation method, because:

-
All six class widths are equal
, at 1515, so a single hh divides every deviation
- The class marks are large and carry decimals, which makes the direct products awkward
- The deviations are all exact multiples of 1515, which they always are when the widths are equal and aa is a class mark

That last point deserves emphasis. If you choose aa to be one of the class marks, then every uiu_i comes out as a whole number — and if you choose aa to be something else, they do not. So choose aa from the class marks, ideally the one with the highest frequency.

Worked example 3 — a distribution where the step-deviation method is not available.
Suppose the classes were 00–1010, 1010–3030, 3030–6060 and 6060–100100, with widths 1010, 2020, 3030 and 4040.

The widths are unequal, so no single hh divides all the deviations into whole numbers, and the step-deviation method cannot be used. Use the assumed mean method instead, which still shrinks the numbers without needing equal widths. The direct method also works, as it always does.

A comparison worth remembering. The three methods relate like this: the direct method has no simplification, the assumed mean method shifts, and the step-deviation method shifts and scales. Each undoes its own simplification at the end, which is why they cannot disagree — and if yours do disagree, one of the columns has an arithmetic error rather than a method error.

How do you find a missing frequency when the mean is given?

Call the missing frequency ff, write both totals as expressions in ff, set the mean formula equal to the given mean, and solve the linear equation.

Worked example 1.
The mean of the following distribution is 1818. Find the missing frequency ff.

- 1111–1313: frequency 33, class mark 1212, product 3636
- 1313–1515: frequency 66, class mark 1414, product 8484
- 1515–1717: frequency 99, class mark 1616, product 144144
- 1717–1919: frequency 1313, class mark 1818, product 234234
- 1919–2121: frequency ff, class mark 2020, product 20f20f
- 2121–2323: frequency 55, class mark 2222, product 110110
- 2323–2525: frequency 44, class mark 2424, product 9696

The two totals as expressions in ff:

∑fi=3+6+9+13+f+5+4=40+f\sum f_i = 3 + 6 + 9 + 13 + f + 5 + 4 = 40 + f
∑fixi=36+84+144+234+20f+110+96=704+20f\sum f_i x_i = 36 + 84 + 144 + 234 + 20f + 110 + 96 = 704 + 20f

Set the mean equal to 1818:

704+20f40+f=18\frac{704 + 20f}{40 + f} = 18
704+20f=18(40+f)=720+18f704 + 20f = 18(40 + f) = 720 + 18f
2f=16⟹f=82f = 16 \quad\Longrightarrow\quad f = 8

Check by computing the mean with f=8f = 8:

∑fi=48,∑fixi=704+160=864\sum f_i = 48, \qquad \sum f_i x_i = 704 + 160 = 864
xˉ=86448=18\bar{x} = \frac{864}{48} = 18

Exactly the given mean. That check is compulsory, because the algebra is easy to get right in a way that still leaves a wrong answer if a product was mis-multiplied.

Notice why the equation came out linear. The unknown ff appears once in the numerator and once in the denominator, and cross-multiplying leaves single powers of ff on both sides. Every missing-frequency question is a one-step linear equation, and if yours turns out quadratic, an ff has been squared by mistake.

Worked example 2 — with the step-deviation method, which is faster here. Solve the same problem taking a=18a = 18 and h=2h = 2, so that ui=xi−182u_i = \frac{x_i - 18}{2}.

- x=12x = 12: u=−3u = -3, fiui=−9f_i u_i = -9
- x=14x = 14: u=−2u = -2, fiui=−12f_i u_i = -12
- x=16x = 16: u=−1u = -1, fiui=−9f_i u_i = -9
- x=18x = 18: u=0u = 0, fiui=0f_i u_i = 0
- x=20x = 20: u=1u = 1, fiui=ff_i u_i = f
- x=22x = 22: u=2u = 2, fiui=10f_i u_i = 10
- x=24x = 24: u=3u = 3, fiui=12f_i u_i = 12

∑fiui=−9−12−9+0+f+10+12=f−8\sum f_i u_i = -9 - 12 - 9 + 0 + f + 10 + 12 = f - 8

Now the mean formula:

18=18+2×f−840+f18 = 18 + 2 \times \frac{f - 8}{40 + f}

so the correction term must be zero:

f−840+f=0⟹f=8\frac{f - 8}{40 + f} = 0 \quad\Longrightarrow\quad f = 8

One line. Because the assumed mean was chosen to equal the given mean, the whole correction had to vanish — so the equation became ∑fiui=0\sum f_i u_i = 0 and nothing else. That is the fastest possible route to a missing frequency, and it works whenever the given mean happens to be one of the class marks.

Worked example 3 — two missing frequencies. If two frequencies are unknown, one equation is not enough. You need a second piece of information, and it is always supplied — usually the total frequency. With ∑fi\sum f_i given as well, you get two linear equations in two unknowns and solve them by elimination, exactly as in the linear-equations chapter.

The general shape of every question in this family. One unknown needs one condition; two unknowns need two. Count your unknowns and count your conditions before you start, and if they do not match, re-read the question — a total frequency or a second mean is hiding in it somewhere.
Exam tip

What does a full-mark mean calculation look like on paper?

Set the work out as a table with a column for every quantity you compute, and total the columns you need. An examiner marks the columns, so working done in your head earns nothing even when the answer is right.

- Make the classes continuous first if they are given with gaps, such as 1111–1313 and 1414–1616
- Write the class mark column as lower+upper2\frac{\text{lower} + \text{upper}}{2}, and check the marks rise by the class width
- Choose aa to be a class mark, preferably the one with the largest frequency, so that every uiu_i is a whole number
- Use the step-deviation method only when all class widths are equal, and say so
- Total the ∑fi\sum f_i column and the ∑fiui\sum f_i u_i column, and show both totals
-
Add aa back and multiply by hh
in the right order: a+h(∑fiui∑fi)a + h\left(\frac{\sum f_i u_i}{\sum f_i}\right), with the hh multiplying only the fraction
- Keep negative deviations negative and add the column carefully — most errors in this chapter are sign errors in the fiuif_i u_i column
- Check the mean lies between the smallest and largest class marks, and on the side the frequencies lean toward
- Verify a missing frequency by recomputing the mean with it substituted
- Name the method you used in your answer

The misconception to name. The mean of grouped data is an estimate, not the exact mean of the original observations. It assumes every value in a class sits at the class mark, which is almost never literally true. The estimate is close because the errors within a class largely cancel, and a question asking "why is this only an estimate?" wants exactly that sentence.

A second trap. Multiplying by hh in the wrong place. The formula is a+h(∑fiui∑fi)a + h\left(\frac{\sum f_i u_i}{\sum f_i}\right), so the hh multiplies the fraction and not the whole expression. Writing h(a+∑fiui∑fi)h\left(a + \frac{\sum f_i u_i}{\sum f_i}\right) scales the assumed mean too, and in this chapter's example would have given 15(47.5+0.967)=72715(47.5 + 0.967) = 727 instead of 6262 — an answer so far out that the plausibility check catches it at once.
Did you know

Why does shifting all the data leave the average shifted by exactly the same amount?

Take any five numbers — say 1212, 1515, 1919, 2222, 2727 — whose mean is 1919. Now subtract 1919 from each: you get −7-7, −4-4, 00, 33, 88. Their mean is 00, since they add to zero.

That is not a property of those particular numbers.
Subtracting the mean from every observation always gives a set whose mean is zero, because the amounts above the mean and the amounts below it are equal by definition — which is what being the mean means.

So the mean is a balance point. Imagine the number line as a see-saw with a unit weight placed at each observation. The mean is the point where it balances
, and the deviations are the distances of the weights from the pivot. The balance condition is exactly ∑(xi−xˉ)=0\sum (x_i - \bar{x}) = 0.

And that is why the assumed mean method works. You pivot the see-saw at aa instead of at the true mean, and measure how far out of balance it is. The quantity ∑fidi∑fi\frac{\sum f_i d_i}{\sum f_i} is exactly how far you must move the pivot to restore the balance — so adding it to aa lands you on the mean.

Which explains two things you may have noticed.

- If aa is chosen below the true mean, ∑fidi\sum f_i d_i comes out positive, and the correction pushes the answer up. In this chapter's example a=47.5a = 47.5 gave +14.5+14.5, and the mean was indeed above aa
- If you happened to choose aa equal to the true mean, the sum would be exactly zero — which is precisely the shortcut that solved the missing-frequency question in one line

The scaling in the step-deviation method has the same kind of explanation.
Measuring distances in units of 1515 instead of 11 divides every deviation by 1515, so the out-of-balance amount comes out in those units too. Multiplying by 1515 converts it back, exactly as converting a distance from metres to centimetres would.

One consequence that matters far beyond this chapter. Because the deviations from the mean always sum to zero, they are useless for measuring spread — you cannot average them to see how scattered the data is, since the answer is always zero. That is why spread is measured by squaring the deviations first, and the quantity that results is the variance. The fact that makes the assumed mean method work is the same fact that forces the definition of variance
, and in Class 11 the two appear a page apart.
Exam relevance

How is the mean of grouped data used in JEE and NEET?

This is foundation work for Class 11 Statistics, which is a JEE Main topic, and the summation notation here is used throughout Physics and Chemistry.

Where the mean formulas lead. Class 11 keeps xˉ=∑fixi∑fi\bar{x} = \frac{\sum f_i x_i}{\sum f_i} unchanged and adds the mean deviation, the variance and the standard deviation, all built on the same grouped table with one extra column. JEE Main asks for the variance or the standard deviation of a grouped distribution as a direct numerical, and the table you learn to lay out here is the table those questions need.

Where the step-deviation idea leads. Class 11 uses exactly the same shift-and-scale trick for variance, with the result that the variance of uiu_i times h2h^2 gives the variance of xix_i. The shift leaves the variance unchanged and the scale multiplies it by h2h^2 — which is the natural next question after seeing the shift and scale behave as they do for the mean, and it is a favourite JEE Main one-liner.

Where the balance-point picture leads. It becomes the centre of mass
in Class 11 Physics, where ∑mixi∑mi\frac{\sum m_i x_i}{\sum m_i} replaces ∑fixi∑fi\frac{\sum f_i x_i}{\sum f_i} with masses in place of frequencies. The formula is identical, and the assumed-mean trick becomes the standard practice of choosing a convenient origin. Both JEE and NEET Physics rely on that choice.

Where the deviations-sum-to-zero fact leads. It is why variance uses squares, and it recurs in Chemistry when averaging bond enthalpies and in error analysis when combining measurements.

Where the missing-frequency technique leads. Competitive papers set problems giving a mean and one unknown observation, or a combined mean of two groups from which one group's mean must be recovered. The method is the same: write the totals as expressions in the unknown and solve.

Question types to expect. At this level: mean by each of the three methods, justifying the choice of method, and a missing frequency. In competitive papers: variance and standard deviation of grouped data, the effect of shifting and scaling on the mean and the variance, and combined means.

The single trap that costs marks. Misplacing the hh in the step-deviation formula. It multiplies the fraction only, and the plausibility check — does the mean lie between the extreme class marks? — catches the error immediately. In Class 11 the corresponding slip is multiplying the variance by hh instead of h2h^2.

A second trap. Using the step-deviation method on unequal class widths. A single hh cannot divide every deviation into whole numbers, so the method does not apply, and a question with widths 1010, 2020, 3030 and 4040 is testing whether you notice. Say which method you are using and why.

Board versus competitive emphasis. The CBSE paper marks the table, each column, the two totals, the substitution and the named method; a competitive paper marks a single value, often a variance rather than a mean. The transferable habit is choosing the origin to make the arithmetic easy — an assumed mean in statistics, a convenient origin in mechanics, a reference level for potential energy. It is the same decision every time, and it always saves work.
Key takeaways

What must you be able to do from this part?

Three methods, one answer and one condition on the fastest of them.

- A class mark is lower limit+upper limit2\frac{\text{lower limit} + \text{upper limit}}{2}, and every observation in a class is treated as sitting there
- The mean of grouped data is therefore an estimate, close because errors within a class largely cancel
- Make classes continuous first if they are given with gaps
- Direct method: xˉ=∑fixi∑fi\bar{x} = \frac{\sum f_i x_i}{\sum f_i}
- Assumed mean method: xˉ=a+∑fidi∑fi\bar{x} = a + \frac{\sum f_i d_i}{\sum f_i}, with di=xi−ad_i = x_i - a
- Step-deviation method: xˉ=a+h(∑fiui∑fi)\bar{x} = a + h\left(\frac{\sum f_i u_i}{\sum f_i}\right), with ui=xi−ahu_i = \frac{x_i - a}{h}
- The hh multiplies the fraction only, not the whole expression
-
All three give the same mean.
For the six-class distribution of 3030 observations they give 186030\frac{1860}{30}, then 47.5+4353047.5 + \frac{435}{30}, then 47.5+15×293047.5 + 15 \times \frac{29}{30} — all equal to 6262
- The sums shrink from 18601860 to 435435 to 2929, which is the whole case for the step-deviation method
-
Choose aa to be a class mark
, ideally the one with the largest frequency, so every uiu_i is a whole number and one product vanishes
- The step-deviation method needs equal class widths — with widths 1010, 2020, 3030, 4040 it cannot be used, and the assumed mean method should be
- For a missing frequency, write both totals in terms of ff and set the mean equal to the given value — a mean of 1818 gives 704+20f40+f=18\frac{704 + 20f}{40 + f} = 18 and so f=8f = 8
- Verify: with f=8f = 8, 86448=18\frac{864}{48} = 18
- If the given mean equals a class mark, take it as aa and the whole correction must vanish, giving ∑fiui=0\sum f_i u_i = 0 in one line
- Two unknown frequencies need two conditions, usually the total frequency as well
- Check the mean lies between the extreme class marks and leans the way the frequencies do
- The deviations from the mean always sum to zero, which is why the mean is the balance point of the data

The most convincing self-test is one table done twice. Take any grouped distribution, find its mean by the direct method and again by the step-deviation method, and see whether the two land on the same number — and notice how much smaller the second set of products was.

Ready to put this into practice?

Create a personalized quiz on this exact topic — free to start.

Create your own quiz on Statistics — Part 1Create a free account
← Back to all articles