Skip to content

Histograms, Distributions, Scatter and Box Plots

CSIR NET Part A: read histograms, ogives, box plots and scatter plots; grouped mean, median and mode, skewness, IQR, outliers, correlation and the normal curve.

  • 18 min read
  • Updated
On this page · 15 sections
  1. Prerequisites
  2. The Hook: Same Average, Different Story
  3. Frequency Tables and Histograms
  4. Mean, Median and Mode from Grouped Data
  5. Shape: Symmetry and Skewness
  6. Box Plots and the Five-Number Summary
  7. Spread and the Normal Curve
  8. Scatter Plots and Correlation
  9. Memory Tricks
  10. Don’t Confuse
  11. Common Mistakes
  12. Practice
  13. Quick Revision Table
  14. Key Takeaways
  15. What’s Next

Prerequisites

The Hook: Same Average, Different Story

Frequency Tables and Histograms

A frequency distribution groups data into classes and counts how many values fall in each. A histogram draws it: classes along the horizontal axis, one bar per class, bars touching because the classes are continuous.

The data below are invented: the time (in minutes) 50 students took to finish a lab quiz.

Time (min)0–1010–2020–3030–4040–50
Students41218106
xychart-beta
    title "Time taken by 50 students"
    x-axis "Time (min)" ["0-10", "10-20", "20-30", "30-40", "40-50"]
    y-axis "Number of students" 0 --> 20
    bar [4, 12, 18, 10, 6]

Conventions to know:

  • Class limits: a value of exactly 20 goes into 20–30, not 10–20 (lower limit included, upper excluded), unless the question says otherwise.
  • Class mark (mid-point): \( \frac{\text{lower} + \text{upper}}{2} \); 25 for the class 20–30.
  • Class width: upper − lower; 10 here.

Unequal class widths: plot frequency density

In a true histogram, area represents frequency. When classes have different widths, the bar height must be the frequency density:

$$\boxed{\text{frequency density} = \frac{\text{frequency}}{\text{class width}}}$$

Frequency polygon and ogive

  • A frequency polygon joins the tops of the bars at their mid-points. It shows the same shape as the histogram as a line, which is handy for overlaying two distributions.
  • An ogive (cumulative frequency curve) plots the running total against the upper class boundary. A “less than” ogive rises from 0 to N; a “more than” ogive falls from N to 0.
Time less than (min)1020304050
Cumulative frequency416344450

Reading an ogive:

  • Median: go across from \( \frac{N}{2} = 25 \) to the curve and down. Here it falls in 20–30.
  • Quartiles: across from \( \frac{N}{4} \) and \( \frac{3N}{4} \).
  • “How many below x”: go up from x to the curve and across.
  • The “less than” and “more than” ogives cross at the median.

The slope of the ogive in any class is that class’s frequency density: the ogive is steepest where the histogram is tallest. This is the “slope of a cumulative curve is the rate” idea from the previous page.

Mean, Median and Mode from Grouped Data

With grouped data you do not know the individual values, so each formula assumes values are spread evenly within a class.

Mean (using class marks \( x_i \) and frequencies \( f_i \)):

$$\boxed{\bar{x} = \frac{\sum f_i x_i}{\sum f_i}}$$

Median (L = lower boundary of the median class, cf = cumulative frequency before it, f = its frequency, h = class width):

$$\boxed{\text{Median} = L + \frac{\frac{N}{2} - cf}{f}\times h}$$

Mode (\( f_1 \) = modal class frequency, \( f_0 \) and \( f_2 \) = the classes before and after):

$$\boxed{\text{Mode} = L + \frac{f_1 - f_0}{2f_1 - f_0 - f_2}\times h}$$

The empirical relation

For moderately skewed distributions, an approximate relation links the three:

$$\boxed{\text{Mode} \approx 3\,\text{Median} - 2\,\text{Mean}}$$

For the quiz data: \( 3(25) - 2(25.4) = 24.2 \), close to the 24.3 computed directly. It is an approximation, not an identity; use it only when a question asks for it or when two of the three are given.

Shape: Symmetry and Skewness

graph LR
    A["Left (negative) skew
long tail to the left
mean lowest, mode highest"] --- B["Symmetric
mean = median = mode"] --- C["Right (positive) skew
long tail to the right
mode lowest, mean highest"]
ShapeOrder of the averagesTypical example
Symmetric, single peakmean = median = modeMeasurement error around a true value
Right (positive) skewmode < median < meanIncomes; time to finish a task; number of citations per paper
Left (negative) skewmean < median < modeMarks in a very easy test (most score high, a few low)
Bimodaltwo peaks; the mean may sit in a valleyTwo groups mixed together (for example, two batches)

Box Plots and the Five-Number Summary

A box plot (box-and-whisker plot) draws five numbers: minimum, first quartile \( Q_1 \), median, third quartile \( Q_3 \), maximum.

  • The box spans \( Q_1 \) to \( Q_3 \) and holds the middle 50% of the data. Its length is the interquartile range, \( \text{IQR} = Q_3 - Q_1 \).
  • The line inside the box is the median.
  • The whiskers reach out to the smallest and largest values that are not outliers.
  • A common outlier rule: a value is an outlier if it lies more than \( 1.5 \times \text{IQR} \) below \( Q_1 \) or above \( Q_3 \).
$$\boxed{\text{Lower fence} = Q_1 - 1.5\,\text{IQR}} \qquad \boxed{\text{Upper fence} = Q_3 + 1.5\,\text{IQR}}$$

Reading a box plot

FeatureMeaning
Median line near the centre of the boxMiddle half roughly symmetric
Median nearer \( Q_1 \); longer right whiskerRight skew
Short boxMiddle half tightly clustered (small IQR)
Points beyond the whiskersOutliers
Two boxes that do not overlapThe middle halves of the two groups do not overlap at all

Two facts follow directly from the definitions: at least 25% of values are at or below \( Q_1 \), and at least 75% are at or above it; about half are above the median.

Spread and the Normal Curve

The range (max − min) uses only two values and is wrecked by one outlier. The IQR ignores the outer quarters and is robust. The standard deviation (SD) uses every value and is the natural measure for symmetric, bell-shaped data.

For data that follow a normal distribution, the empirical rule gives:

IntervalShare of values (approximately)
mean ± 1 SD68%
mean ± 2 SD95%
mean ± 3 SD99.7%

Changing every value by the same amount shifts the mean and median by that amount but leaves the SD, IQR and range unchanged. Multiplying every value by \( k \) (with \( k > 0 \)) multiplies all of them by \( k \).

Scatter Plots and Correlation

A scatter plot shows one dot per individual, with one variable on each axis. Read three things:

  1. Direction: dots rising to the right (positive association) or falling (negative).
  2. Form: a straight-line band, a curve, or no pattern.
  3. Strength: how tightly the dots hug the line or curve.

The correlation coefficient \( r \) measures the strength and direction of a linear relation.

\( r \)What the scatter looks like
+1All points exactly on a rising straight line
about +0.8Clear rising band, some scatter
about 0No linear trend (there may still be a curved one)
about −0.8Clear falling band
−1All points exactly on a falling straight line

Facts to have ready:

  • \( -1 \le r \le 1 \), and \( r \) has no units.
  • \( r^2 \) is the fraction of the variation in y accounted for by the straight-line fit. \( r = 0.8 \) gives \( r^2 = 0.64 \): 64%, not 80%.
  • The sign of \( r \) is the sign of the slope of the best-fit line; the size of \( r \) is not the slope.
  • \( r \) is unchanged if you change units by a positive factor (cm to m) or add a constant to every value; multiplying one variable by a negative number flips its sign.
  • One outlier can raise or lower \( r \) a lot.

Correlation is not causation

A strong correlation between X and Y can arise because:

  • X causes Y;
  • Y causes X;
  • a third variable drives both (a confounder: ice-cream sales and drowning deaths both rise in summer);
  • coincidence, especially when many pairs of variables are tried.

A Part A statement such as “the scatter plot proves that X causes Y” is false whatever the plot shows. “X and Y are positively associated in this sample” is the most a scatter plot alone supports. For the formal theory of correlation and regression, see the statistics side of Mathematical Sciences; Part A needs only the reading skills on this page.

Memory Tricks

Don’t Confuse

ThisvsThat
Bar chart (separate categories, gaps between bars)histogram (continuous classes, bars touch, area = frequency)
Frequency (count in a class)frequency density (count ÷ width; the bar height for unequal widths)
Cumulative frequency before the median class (used in the formula)cumulative frequency up to and including it
Range (max − min)IQR (\( Q_3 - Q_1 \))
\( r = 0.8 \)80% explained (it is \( r^2 = 64\% \))
Strong correlationcausation
\( r = 0 \)no relationship of any kind

Common Mistakes

Practice

All questions below are original practice questions written for this page. They are not from any CSIR NET paper. Data sets are invented.

Level 1: Foundation

  1. What is plotted on the vertical axis of a “less than” ogive?
  2. In a distribution with a long tail to the right, order the mean, median and mode.
  3. Define the interquartile range.
Answers
  1. Cumulative frequency (the number of values less than each upper class boundary).
  2. Mode < median < mean.
  3. IQR = \( Q_3 - Q_1 \), the spread of the middle 50% of the data.

Level 2: Part A Style

Practice Q1. Twenty seedlings are grouped by height (cm): 0–10: 3, 10–20: 7, 20–30: 5, 30–40: 5. Using class mid-points, the mean height is:

(a) 19 cm (b) 20 cm (c) 21 cm (d) 22 cm

Answer: (c) 21 cm

\( \sum f x = 3(5) + 7(15) + 5(25) + 5(35) = 15 + 105 + 125 + 175 = 420 \). Mean \( = \frac{420}{20} = 21 \) cm.

Practice Q2. In a histogram drawn correctly, the class 20–40 has frequency 30 and the class 40–50 has frequency 25. The ratio of the height of the 20–40 bar to the height of the 40–50 bar is:

(a) 6 : 5 (b) 3 : 5 (c) 5 : 3 (d) 12 : 5

Answer: (b) 3 : 5

Heights are frequency densities: \( \frac{30}{20} = 1.5 \) and \( \frac{25}{10} = 2.5 \). Ratio \( 1.5 : 2.5 = 3 : 5 \).

Option (a) compares raw frequencies, which is wrong because the widths differ.

Practice Q3. A scatter plot of hours of instrument time booked against number of samples processed, for 30 research groups, gives \( r = -0.9 \). Which statement is best supported?

(a) Booking more instrument time causes groups to process fewer samples. (b) Groups that booked more time tended to process fewer samples, and the relation is close to linear. (c) 90% of the variation in samples processed is explained by instrument time. (d) There is no relation, because r is negative.

Answer: (b)

\( r = -0.9 \) is a strong negative linear association. (a) claims causation, which a correlation cannot show. (c) confuses \( r \) with \( r^2 \) (here \( r^2 = 0.81 \), so about 81%). (d) confuses a negative relation with no relation.

Practice Q4. Scores of 400 candidates are approximately normal with mean 70 and standard deviation 5. The number of candidates scoring above 80 is closest to:

(a) 2 (b) 10 (c) 20 (d) 64

Answer: (b) 10

80 is 2 SD above the mean. About 95% lie within ±2 SD, so about 2.5% lie above 80: \( 0.025 \times 400 = 10 \).

Option (c) uses 5% (both tails) instead of 2.5% (one tail). Option (d) uses the 1 SD tail (16%) for a 2 SD question.

Level 3: Multi-Step (Part C depth)

Practice Q5. A “less than” cumulative frequency table for the masses (g) of 50 fruits reads:

Less than (g)1020304050
Cumulative frequency515324450

The median mass is closest to:

(a) 24.1 g (b) 25.0 g (c) 25.9 g (d) 27.4 g

Answer: (c) 25.9 g

Class frequencies: 5, 10, 17, 12, 6. \( \frac{N}{2} = 25 \) falls in 20–30 (cumulative 15 before it, 32 after).

Median \( = 20 + \frac{25 - 15}{17}\times 10 = 20 + \frac{100}{17} \approx 25.88 \) g.

Using the cumulative frequency including the class (32) instead of 15 gives \( 20 + \frac{25 - 32}{17}\times 10 \approx 15.9 \), below the class’s own lower limit: a sure sign of the wrong cf.

Practice Q6. Box plots of marks for two batches give these five-number summaries:

BatchMin\( Q_1 \)Median\( Q_3 \)Max
A2035455570
B3040505290

Consider the statements:

  1. Batch B has the higher median.
  2. Batch B has the smaller interquartile range.
  3. By the \( 1.5 \times \text{IQR} \) rule, batch B’s maximum is an outlier.
  4. More than half of batch A scored above 50.

Which statements are correct?

(a) 1 and 2 only (b) 1, 2 and 3 only (c) 2 and 3 only (d) 1, 2, 3 and 4

Answer: (b) 1, 2 and 3 only
  • Medians: 50 against 45. 1 true.
  • IQR: A \( = 55 - 35 = 20 \); B \( = 52 - 40 = 12 \). 2 true.
  • B’s upper fence \( = 52 + 1.5 \times 12 = 70 \). The maximum, 90, is above it. 3 true.
  • A’s median is 45, so at least half of A scored 45 or less; at most half scored above 45, so more than half cannot be above 50. 4 false.

Note that B also has the larger range (60 against 50) despite its smaller IQR: the range is driven by one extreme value.

Practice Q7. Twelve readings are 12, 15, 17, 18, 20, 22, 23, 25, 28, 30, 31 and 50. The reading 50 is found to be a recording error and is removed. Which statement is correct?

(a) Both the mean and the median fall by about 2.3. (b) The mean falls by about 2.3; the median falls by 0.5. (c) The mean falls by 0.5; the median falls by about 2.3. (d) The mean falls by about 2.3; the median is unchanged.

Answer: (b)
  • Mean with 50: \( \frac{291}{12} = 24.25 \). Without: \( \frac{241}{11} \approx 21.91 \). Fall \( \approx 2.34 \).
  • Median with 50 (12 values): \( \frac{22 + 23}{2} = 22.5 \). Without (11 values): the 6th value, 22. Fall \( = 0.5 \).

The median moves only because the count changes from even to odd; it is robust to the size of the outlier. The mean absorbs the whole error.

Quick Revision Table

IdeaRule
HistogramArea = frequency; unequal widths → height = frequency ÷ width
OgiveCumulative frequency vs upper boundary; median at \( \frac{N}{2} \)
Grouped mean\( \frac{\sum f x}{\sum f} \) with class marks
Grouped median\( L + \frac{N/2 - cf}{f}h \), cf before the class
Grouped mode\( L + \frac{f_1 - f_0}{2f_1 - f_0 - f_2}h \)
Empirical relationMode ≈ 3 median − 2 mean
SkewRight: mode < median < mean; left: reverse
Box plotBox = \( Q_1 \) to \( Q_3 \); IQR = box length
Outlier fences\( Q_1 - 1.5\,\text{IQR} \), \( Q_3 + 1.5\,\text{IQR} \)
Normal curve68–95–99.7% within 1, 2, 3 SD
Correlation\( -1 \le r \le 1 \); \( r^2 \) = share explained; not causation

Key Takeaways

What’s Next

Advertisement
Advertisement