Chapter 15 — Statistics

What is Statistics? Statistics is the branch of mathematics that deals with collection, organization, presentation, analysis and interpretation of numerical data to make decisions. It converts raw data into useful information through suitable methods.

Objectives / Steps in a Statistical Study

• Collection of data (primary or secondary)
• Organization of data (classification and tabulation)
• Presentation of data (graphs, charts, tables)
• Analysis and interpretation (measures of central tendency, dispersion, etc.)
• Drawing conclusions and making decisions

Types of Data

• Qualitative (Categorical): Non-numeric attributes (e.g., blood group, city names).
• Quantitative (Numerical): Numeric values. Further divided into:
  • Discrete (countable values: number of students)
  • Continuous (measurable values: height, weight, temperature)

Primary vs Secondary Data: Primary data are obtained by direct observation or survey. Secondary data are collected from published sources or other studies.

Classification and Frequency Distribution: For quantitative data (especially large data), we group into class intervals and make a frequency distribution giving class limits, class boundaries, class width (h), class mark (mid-point) and frequency (f).

Key Terms

• Class width (h): upper limit − lower limit of a class
• Class mark (mid-point) x_i: (lower limit + upper limit) / 2
• Frequency (f_i): number of observations in a class
• Cumulative frequency (cf): running total of frequencies up to a class
• Relative frequency: f_i / n (often expressed as a fraction or percentage)

Presentation of Data: Data are presented using tables and graphs for clarity. Common graphs: bar charts (for categorical/discrete data), histograms and frequency polygons (for continuous/grouped data), pie charts (for categorical proportions), and ogives (cumulative frequency curves).

Uses and Limitations: Statistics helps summarize large data sets and make informed decisions (policy, business, science). Limitations include sensitivity to biased sampling, poor data collection, and misleading graphics if used improperly.

What is data? Data are raw facts and figures collected for analysis. In Statistics, collecting data is the first step: you record observations about a variable(s) from a population or a sample.

Types of data

• Qualitative (Attribute): non-numeric categories (e.g., blood group, colour).
• Quantitative: numeric measurements.
  • Discrete: countable values (e.g., number of students).
  • Continuous: measurable on a continuum (e.g., height, weight).

Scales of measurement: nominal, ordinal, interval, ratio. These determine what analyses and graphs are appropriate.

Sources of data

• Primary data: collected first-hand for the study (surveys, experiments, observations).
• Secondary data: obtained from existing sources (books, articles, government reports).

Methods of collecting primary data

• Observation: record what is seen (useful when respondents can’t report reliably).
• Questionnaire / Interview: ask participants via written forms or oral interviews.
• Schedule: structured form filled by investigator after interviewing the respondent.
• Experiment / Measurement: controlled conditions, often for scientific studies.

Sampling methods (brief): simple random sampling, stratified sampling, systematic sampling, cluster sampling, convenience sampling. Choose based on study goals, resources and need for representativeness.

Classification and tabulation: Raw data are grouped into classes (for continuous data) or categories (for discrete/qualitative data) to make frequency distributions (tables of class/cateogry vs frequency). Steps: 1) Decide variable type and scale, 2) Choose number of classes (k), 3) Determine class width/intervals, 4) Tally frequencies, 5) Compute relative frequencies/cumulative frequencies if needed.

Coding data: Often raw values are transformed by a linear change to simplify calculations: y = (x − a)/b (shift by a and scale by b). Recover original by x = b*y + a.

Important practical points

• Design questionnaires to avoid ambiguity and bias; pilot-test them.
• Ensure sample represents the population if generalisation is required.
• Watch for non-response and measurement errors; document data source and method.

From collection to visualization: After classifying and tabulating, data are summarized with graphs (bar charts, histograms, pie charts, ogives, scatter plots). Choice of graph depends on data type: qualitative → bar/pie; quantitative (continuous) → histogram, frequency polygon, ogive; bivariate quantitative → scatter plot.

Definition: Classification is the process of arranging raw data into groups or classes according to some rule. Tabulation is the systematic presentation of classified data in the form of tables showing frequencies and other summary measures.

Types of data (brief):

• By source: Primary (collected first-hand), Secondary (from existing records).
• By nature: Qualitative (categorical) and Quantitative (numerical).
• Quantitative can be Discrete (take specific values, e.g., number of children) or Continuous (measured, e.g., height, weight).
• Measurement scales (useful to know): Nominal, Ordinal, Interval, Ratio.

Why classify and tabulate? To simplify large data sets, reveal patterns, allow computation of statistical measures (mean, median, mode, dispersion), and prepare data for graphical presentation.

Key terms:

• Class interval: a range of values (e.g., 30–39).
• Class limits: lower and upper limits of the interval.
• Class boundaries: exact end points when intervals are made continuous (e.g., 29.5–39.5 for integer data).
• Class width (h): difference between upper and lower limits of a class (constant for equal-width classes).
• Class mark (midpoint) xi: (lower limit + upper limit) / 2.
• Frequency fi: number of observations in a class.
• Relative frequency: fi / N (N = total observations).
• Cumulative frequency: running total of frequencies up to (or from) a class.

Steps to create a grouped frequency distribution:

1. Arrange raw data in ascending order (optional but helpful).
2. Decide number of classes k. Common rules: 1 + 3.3 log10 n (Sturges' rule) or k roughly between 5 and 15; sometimes k ≈ √n.
3. Compute class width h ≈ (max − min) / k and round to a convenient value.
4. Choose the first lower limit; build k mutually exclusive, exhaustive class intervals of equal width.
5. Tally observations into classes and obtain frequencies fi.
6. Compute class marks, relative frequencies, percentage frequencies and cumulative frequencies as needed.

Guidelines for classes:

• Classes should not overlap and must cover the entire range.
• Prefer equal class widths for easier interpretation and graphical work (histogram, frequency polygon).
• For discrete integer data, use boundaries like 9.5–19.5 to avoid gaps when drawing histograms.
• Use open-ended classes only when necessary (e.g., "60 and above").

Inclusive vs exclusive class limits: In inclusive form (used for integer data) an interval 10–14 means values 10,11,12,13,14. For continuous data or for drawing histograms, use class boundaries (e.g., 9.5–14.5).

Common mistakes to avoid:

• Overlapping classes or leaving gaps.
• Unequal or odd class widths without reason when comparison is required.
• Incorrect cumulative counting direction for ogives (be clear whether you use "less than" or "more than" ogive).

Summary: Classification groups observations into meaningful categories; tabulation arranges these groups neatly with frequencies and related measures so that further statistical analysis and graphical representation become easy and reliable.

What is presentation of data? Presentation of data means arranging and displaying collected numerical or categorical observations in tables and graphs so that important features (patterns, trends, comparisons) become clear and easy to interpret.

Why present data? To summarise large data sets, reveal patterns (central tendency, spread, skewness), enable comparisons, and support decision making.

Types of data: qualitative (categorical) and quantitative (numerical). Quantitative can be discrete (countable) or continuous (measured).

Tabular presentation — frequency distribution. Two main kinds:

• Ungrouped frequency distribution: list distinct values and their frequencies (used when number of distinct values is small).
• Grouped frequency distribution: data grouped into class intervals (used for large/continuous data).

Steps to make a grouped frequency distribution (typical):

1. Arrange raw data; find n (total observations), minimum and maximum.
2. Compute range R = max − min.
3. Choose number of classes k (common rules: 5–15; sometimes k ≈ √n).
4. Class width (approx) h = R / k (round up to convenient number).
5. Form k consecutive non-overlapping class intervals of width h covering all data.
6. Tally frequencies fi for each class; compute cumulative frequencies (CF), class marks xi, relative frequencies (fi/n) and percentages.

Important terms:

• Class limits: lower and upper limits of a class.
• Class boundaries: adjust limits by half the least measurement unit to remove gaps (e.g., if data are integers, class 10–19 and 20–29 have boundaries 9.5–19.5 and 19.5–29.5).
• Class mark (midpoint): xi = (lower limit + upper limit) / 2.
• Cumulative frequency (CF): running total of frequencies up to a class.
• Relative frequency: fi / n; percentage = (fi / n) × 100%.

Graphical presentation — choose depending on data and purpose:

• Bar diagram: categorical or discrete data comparisons (gaps between bars).
• Histogram: continuous data grouped into classes (contiguous bars; area of bar ∝ frequency; height = frequency / class width for unequal widths).
• Frequency polygon: plot class marks against frequencies and join by straight lines (useful to compare distributions).
• Ogive (cumulative frequency curve): plot upper class limits (or boundaries) vs cumulative frequency; useful to read medians/percentiles.
• Pie chart: show parts of a whole for categorical data (use when number of categories is small).
• Pictogram/stem-and-leaf: alternative visual summaries (pictogram for simple categories; stem-and-leaf retains raw-data information).

Rules & tips:

• Choose class width and k so classes are neither too few (loss of detail) nor too many (noisy).
• Use equal class widths if possible; if unequal widths are necessary, draw histogram with bar area proportional to frequency (height = frequency / width).
• Label axes, give title, and include units and legends where appropriate.

Graphical Representation of Data is the use of diagrams and graphs to present statistical data visually so patterns, trends and comparisons become clear at a glance. It is a key part of Statistics in Class 11 because it helps interpret raw and grouped data without lengthy calculations.

Major types of graphs and when to use them:

• Bar graph — for qualitative (categorical) data or discrete numerical data (gaps between bars).
• Pie chart — to show percentage/share of a whole for categories (useful for market-share, budget splits).
• Histogram — for continuous grouped data; contiguous bars show frequency per class interval. If class widths differ, use frequency density (frequency/width) for bar height.
• Frequency polygon — join class-midpoint frequencies with straight lines; useful to compare distributions and show smoothing of a histogram.
• Frequency curve — smooth curve drawn through points of a frequency polygon (approximate continuous distribution).
• Ogive (cumulative frequency curve) — plots cumulative frequency vs class boundary; two types: ‘less than’ and ‘greater than’. Useful for reading medians, quartiles and percentiles graphically.
• Scatter plot (XY-plot) — for bivariate data to examine relationship (correlation/causation) between two variables.
• Time-series (line graph) — for data indexed by time to show trends, seasonal patterns.

General steps to construct common graphs:

• Histogram: determine class intervals and class boundaries, compute frequencies (or frequency density if widths unequal), draw contiguous rectangles with base equal to class width and height equal to frequency or frequency density.
• Frequency polygon: compute class midpoints, plot midpoint vs frequency, join successive points with straight lines; close polygon by joining first and last points to the horizontal axis at the class-endpoints.
• Ogive: compute cumulative frequencies (less-than: cumulative up to upper class boundary); plot boundary vs cumulative frequency and join by a smooth/straight curve.
• Pie chart: convert each frequency to angle = (frequency/total)×360°, draw sectors accordingly.

Interpreting graphs: identify central tendency (peak of histogram or polygon), spread (width of distribution), skewness (tail left/right), modality (uni-/bi-/multi-modal), outliers, and trends (in time series or scatter plots).

Advantages: quick visual comparison, intuitive pattern recognition. Limitations: can hide exact values, depend on choice of class width and origin (histogram) or angle rounding (pie chart).

Cumulative Frequency

Cumulative frequency (CF) for a class or value is the running total of frequencies up to that class/value. If f1, f2, ..., fk are frequencies of successive classes, then the cumulative frequencies are CF1 = f1, CF2 = f1 + f2, ..., CFk = f1 + f2 + ... + fk. Cumulative frequency organizes data to show how many observations lie below (or above) a given value.

Types of Cumulative Frequency

• Less-than cumulative frequency: CF at an upper class boundary = total frequency of all classes with values ≤ that upper boundary. Used to make the "less-than" ogive.
• Greater-than cumulative frequency: CF at a lower class boundary = total frequency of all classes with values ≥ that lower boundary. Used to make the "greater-than" ogive. It can be computed as (Total N) − (less-than CF before that boundary).

Preparing data (grouped data)

1. Start with a frequency distribution table of class intervals and frequencies.
2. Decide correct class boundaries (for continuous data use the class limits; for integer data adjust boundaries by 0.5 if classes are written as whole numbers: e.g. 10–19 implies boundaries 9.5–19.5).
3. Compute cumulative (less-than) frequencies using upper class boundaries. Optionally compute greater-than CF using lower boundaries.

Ogives (Cumulative Frequency Curves)

An ogive is a graph of cumulative frequency versus class boundary/value.

• Less-than ogive: Plot points (upper class boundary, cumulative frequency up to that boundary). Start at the lower-most boundary with CF = 0 (optional) and end at the highest boundary with CF = N.
• Greater-than ogive: Plot points (lower class boundary, cumulative frequency of values greater than or equal to that boundary). It typically starts at N and ends at 0.

Reading information from ogives

• Median: draw a horizontal line at N/2 on the cumulative-frequency axis and read the corresponding x-value on the ogive. If you draw both less-than and greater-than ogives, their intersection gives the median.
• Quartiles/percentiles: draw horizontal lines at N/4, N/2, 3N/4 (or other percentile positions) and read the corresponding x-values on the ogive.
• Proportion below/above a value: read CF at that value and divide by N to get proportion or percentage.

Why ogives are useful

Ogives provide a clear picture of cumulative distribution: they show how quickly observations accumulate with increasing value, enable quick estimation of medians and percentiles, and help compare distributions.

What are Measures of Central Tendency?

Measures of central tendency are single values that describe the centre or typical value of a data set. They summarise a distribution by identifying a central point around which data cluster. The three commonly used measures are Arithmetic Mean, Median and Mode.

1. Arithmetic Mean (Average)

• Definition: Sum of all observations divided by the number of observations.
• Use: Best for quantitative data without extreme outliers; widely used in science, finance and education.
• Properties: Unique, uses all values, sensitive to extreme values (not robust).

2. Median

• Definition: Middle value when observations are arranged in ascending order. For even n, median is the average of the two middle values.
• Use: Preferred for skewed data (e.g. incomes, house prices) because it is not affected by extreme values.
• Properties: Robust (resistant to outliers), may not use all data values.

3. Mode

• Definition: Value that occurs most frequently in the data set. A distribution may be unimodal, bimodal, or multimodal.
• Use: Useful for categorical data (e.g. most common shoe size, favourite colour) and for identifying peaks in a distribution.
• Properties: Can be non-unique; simple to identify for qualitative data.

Grouped Data vs Raw Data

For raw (individual) data the mean, median and mode are computed directly from the list of numbers. For grouped data (data in class intervals) we use formulas and approximations:

• Grouped mean: use class mid-points and frequencies; or use the assumed mean method to simplify calculation.
• Grouped median: approximate position n/2 and interpolate inside the median class using class width and cumulative frequency.
• Grouped mode: identify the modal class (highest frequency) and use formula to approximate the mode by interpolation inside that class.

Relation and Skewness

In practice, the relative position of mean, median and mode indicates skewness:

• Symmetric distribution: Mean ≈ Median ≈ Mode.
• Right (positive) skew: Mode < Median < Mean.
• Left (negative) skew: Mean < Median < Mode.

Choosing the Appropriate Measure

• Use mean for balanced quantitative data without outliers and when mathematical manipulation is needed.
• Use median for skewed distributions or when a resistant measure is required (e.g. income data).
• Use mode for categorical data or to identify the most frequent category/value.

Summary

Mean, median and mode give complementary information about the centre of a distribution. Understanding their definitions, calculation methods (raw and grouped), properties and behavior under skewness helps in choosing the most meaningful measure for a given real-life situation.

Definition: The arithmetic mean (often called simply the mean) of a data set is the sum of all observations divided by the number of observations. It is a measure of central tendency that represents the typical value of the data.

When to use: Use the arithmetic mean for quantitative (numerical) data — both ungrouped (raw) data and grouped (binned) data. It is best when data have no extreme outliers and when every value contributes equally.

Formulas & key ideas:

• Ungrouped data: mean x̄ = (Σx)/n, where Σx is the sum of observations and n is the number of observations.
• Grouped data (with class marks): x̄ = (Σ f x_i)/Σ f, where f is class frequency and x_i is the class mark (midpoint) x_i = (lower limit + upper limit)/2.
• Assumed-mean method (useful for grouped data): choose an assumed mean A, let d_i = x_i − A, then x̄ = A + (Σ f d_i)/Σ f.
• Step-deviation method (reduces big numbers): let h be class width and d'_i = (x_i − A)/h, then x̄ = A + h·(Σ f d'_i)/Σ f.
• Combined mean of two groups: x̄ = (n1 x̄1 + n2 x̄2)/(n1 + n2).
• Important properties: Σ(x_i − x̄) = 0 and Σ x_i = n x̄. For linear transformation y = a x + b, mean(y) = a·mean(x) + b.

Limitations: The mean is sensitive to extreme values (outliers) and may not represent a skewed distribution as well as the median.

How to compute (steps): For raw data, add all values and divide by count. For grouped data, replace each class by its class mark, multiply each class mark by the class frequency, sum these products, and divide by total frequency. For large grouped data, use assumed-mean or step-deviation to simplify arithmetic.

Definition: The median of a dataset is the value that separates the ordered data into two equal parts — 50% of observations are less than or equal to it and 50% are greater than or equal to it. It is the 50th percentile and a measure of central tendency that is robust to extreme values.

General steps to find the median:

1. Arrange the data in ascending order.
2. Find the middle position (depends on the number of observations).
3. If needed (for grouped data), use interpolation inside the median class.

1. Ungrouped (raw) data:

• If n (number of observations) is odd: median is the value at position (n + 1)/2 in the ordered list.
• If n is even: median is the average of values at positions n/2 and (n/2) + 1.

2. Discrete frequency distribution (individual values with frequencies):

• Compute cumulative frequencies. Let N be total frequency. Locate the value (or class) where cumulative frequency ≥ N/2. That value (or category) is the median (or the median class if values are grouped by categories).

3. Continuous grouped frequency distribution (class intervals with frequencies):

• Find total frequency N and cumulative frequencies up to each class.
• Identify the median class: the first class whose cumulative frequency ≥ N/2.
• Use linear interpolation inside the median class (assumes uniform distribution within the class) with the formula: Median = l + ((N/2 - cfbefore) / f_med) * h
• Here, l = lower boundary of the median class (use class boundaries, not limits), cfbefore = cumulative frequency before the median class, f_med = frequency of the median class, h = class width.

Important properties:

• Median is unaffected by extreme outliers (robust).
• For symmetric distributions mean = median = mode (approximately).
• Median minimizes the sum of absolute deviations: it is the value m that minimizes sum |x_i - m|.
• Median is the 50th percentile (P50).

Worked numeric examples (briefly explained):

1. Ungrouped odd n: Data = {12, 15, 11, 20, 18}. Sorted = {11,12,15,18,20}, n=5, position (5+1)/2 = 3 → Median = 15.
2. Ungrouped even n: Data = {10,20,30,40}. Sorted same, n=4, positions 2 and 3: (20+30)/2 = 25 → Median = 25.
3. Grouped continuous example: Classes: 0–10 (5), 10–20 (7), 20–30 (12), 30–40 (6). N = 30, N/2 = 15. Cumulative freqs: 5,12,24,30 → median class is 20–30 (cfbefore = 12, f_med = 12, l = 20, h = 10). Median = 20 + ((15 - 12)/12)*10 = 20 + (3/12)*10 = 20 + 2.5 = 22.5.

Definition: The mode of a data set is the value (or values) that occur most frequently. It is the observation with the maximum frequency. Mode applies to numerical and categorical data.

Types: unimodal (one mode), bimodal (two modes), multimodal (more than two modes), or no mode (all values occur equally).

Ungrouped (raw) data: For a list of observations, identify the value(s) with the highest count. Example: in {2, 3, 5, 3, 8, 3, 5}, the mode is 3 because it appears most often.

Grouped (continuous) data: For a frequency distribution with class intervals, find the modal class (the class with largest frequency). The mode is estimated by interpolation using the formula:

Mode ≈ l + [(f_m - f_{1}) / (2f_m - f_{1} - f_{2})] × h

where l = lower limit of modal class, f_m = frequency of modal class, f_{1} = frequency of the class before modal class, f_{2} = frequency of the class after modal class, and h = class width (length of modal class interval).

Notes / Cautions:

• If class widths are unequal, compare frequency densities (frequency ÷ class width) to choose the modal class; the standard interpolation formula assumes equal widths.
• Mode is the only measure of central tendency applicable to categorical (nominal) data (e.g., most common colour).
• Mode can be less stable than mean/median for small samples but is intuitive and simple to find.

Procedure to compute modal value for grouped data: identify modal class → note l, f_m, f_{1}, f_{2}, h → substitute into the formula → calculate the estimated mode.

Relation with other measures: Mode, median and mean are different summary measures; for symmetric unimodal distributions they coincide, but in skewed distributions the mode is the peak (most frequent value) and may be quite different from mean/median.

Definition: Measures of position locate a particular value in a data set relative to other observations. Common measures are quartiles, deciles and percentiles (including the median as the 50th percentile). They answer questions such as “What value separates the lowest 25% from the rest?”

Key ideas:

• Order the data from smallest to largest for ungrouped (raw) data.
• A percentile P (0<P<100) is the value below which P% of the observations lie. Quartiles and deciles are special percentiles: Q1 = 25th percentile, Q2 = 50th (median), Q3 = 75th; decile Dj = j*10th percentile.
• If the computed position/index is not an integer, interpolate between neighbouring observations (for raw data) or use linear interpolation inside the class (for grouped data).

Ungrouped data (discrete list) — procedure and interpolation:

• Order values x1 ≤ x2 ≤ ... ≤ xn.
• Position (index) for the Pth percentile: i = (P/100)·(n + 1). For quartiles: Q1 at (n+1)/4, Q2 at (n+1)/2, Q3 at 3(n+1)/4. For decile Dj: position j(n+1)/10.
• If i is integer, the value at that index is the percentile. If i = k + d (k integer, 0<d<1), percentile = x_k + d·(x_{k+1} − x_k).

Grouped data (continuous class intervals):

Find the class that contains the required cumulative frequency position (kN/100). Then use linear interpolation inside that class:

Value = L + [ (kN/100 − c.f.) / f ] · h

where L = lower boundary of the class containing the percentile, c.f. = cumulative frequency before that class, f = frequency of that class, h = class width, N = total frequency, k = percentile number (e.g., k = 25 for Q1, k = 90 for 90th percentile).

Interpretation: Saying a student is at the 85th percentile means they scored better than 85% of the group. Quartiles summarize location and spread and are used in boxplots to show central tendency and variability.

Overview: This topic collects the key formulas and standard procedures used to summarise and analyse univariate statistical data (both ungrouped and grouped). It explains how to compute measures of central tendency (mean, median, mode), dispersion (variance, standard deviation), and practical techniques (assumed mean, step-deviation, interpolation for median, modal-class formula). The emphasis is on reliable step-by-step procedures you can apply to real datasets.

Key concepts and notation:

• x or x_i : observed values (raw data)
• f_i : frequency of the i-th class or value
• N = Σ f_i : total frequency (sample size)
• c_i : class boundaries or class intervals; h : class width (for equal-width classes)
• m_i : class-mark (midpoint) = (lower bound + upper bound)/2
• A : assumed mean (a convenient class mark used to simplify arithmetic)
• u_i : step-deviation = (m_i − A)/h (usually an integer)

Worked procedures (step-by-step):

• Arithmetic mean — Ungrouped data: sort data (optional), compute Σx and N, mean x̄ = Σx/N.
• Arithmetic mean — Grouped data (class intervals): compute class-marks m_i, then x̄ = Σ(f_i m_i)/N. If numbers are large, choose an assumed mean A and use step-deviation: x̄ = A + h*(Σ f_i u_i)/N, where u_i = (m_i − A)/h.
• Median — Ungrouped data: sort data and pick middle value (if N odd) or average of two middle values (if N even).
• Median — Grouped data (interpolation): find cumulative frequencies and locate the median class (where cumulative frequency ≥ N/2). Use interpolation formula: Median = L + ((N/2 − c_f)/f_m) * h, where L = lower boundary of median class, c_f = cumulative frequency of preceding class, f_m = frequency of median class, h = class width.
• Mode — Ungrouped data: the value with the highest frequency (if multimodal, there may be more than one).
• Mode — Grouped data (modal class interpolation): identify modal class (class with largest f). Use formula: Mode = L + ((f_1 − f_0)/(2f_1 − f_0 − f_2)) * h, where f_1 is frequency of modal class, f_0 previous class frequency, f_2 next class frequency, L is lower boundary of modal class, h is class width.
• Variance & Standard Deviation — Grouped/ungrouped: variance σ^2 = (1/N) Σ f_i (x_i − x̄)^2. Standard deviation σ = √σ^2. Use step-deviation to simplify: σ^2 = h^2 * [Σ f_i u_i^2 / N − (Σ f_i u_i / N)^2].
• Combined (pooled) mean and variance: For two groups of sizes N1, N2 with means x̄1, x̄2: combined mean x̄ = (N1 x̄1 + N2 x̄2)/(N1 + N2). For variance, use: combined σ^2 = [N1(σ1^2 + x̄1^2) + N2(σ2^2 + x̄2^2)]/(N1+N2) − x̄^2.

Practical tips & techniques:

• Always check if classes are continuous (use class boundaries) and of equal width; if not equal, use actual widths in formulas and class-marks carefully.
• Choose an assumed mean A near the center to make u_i small integers — reduces arithmetic errors.
• For median interpolation, use cumulative frequencies (either ‘less than’ ogive or cumulative Σf) to locate N/2 quickly.
• When using modal formula, ensure modal class is clearly the highest frequency; if adjacent classes tie, mode may be ambiguous.
• Report final answers with appropriate rounding and units (marks, rupees, kg, etc.).

When to use which method:

• Use raw formulas for small ungrouped datasets.
• Use grouped formulas and interpolation when data are presented in class intervals (large datasets).
• Use step-deviation for grouped mean/variance when class-marks are large or spread widely.

What this topic covers: How numerical summaries and graphs (mean, median, mode, dispersion, correlation, regression, and plots) are used to describe real data, draw conclusions, make comparisons and inform decisions. It also covers common pitfalls in interpreting results (bias, outliers, correlation vs causation) and how to present results so they are meaningful and honest.

Why it matters: Statistics turn raw data into actionable information. For example, policy makers, businesses and researchers use averages, variability and relationships to decide resource allocation, set prices, evaluate programs and make predictions.

Key ideas and interpretation:

• Central tendency (mean, median, mode): indicate a typical value. Choose median when data are skewed or contain outliers; mean is useful for further algebraic work (e.g., total/averages).
• Dispersion (range, interquartile range, variance, standard deviation, coefficient of variation): show how spread out values are. High dispersion means more variability and less predictability.
• Shape and outliers: skewness (right/left) affects relationship between mean and median. Outliers can distort the mean but have less effect on the median and IQR.
• Association (correlation and regression): correlation measures strength and direction of linear association; regression provides a predictive linear model. Remember: correlation ≠ causation.
• Sampling and bias: ensure samples are representative. Non-random sampling, response bias, or measurement error can lead to misleading summaries.
• Comparisons: use coefficient of variation (CV) to compare relative variability of datasets with different units or scales.
• Communicating results: choose clear graphs and report context (sample size, units, measures used). Avoid cherry-picking or mis-scaling axes that mislead.

How to interpret typical outputs:

• Mean ± standard deviation (e.g., 50 ± 8): roughly describes the typical spread if distribution is near normal.
• Median and IQR (e.g., median = 48, IQR = 10): gives the middle and spread of the central 50% and is robust to outliers.
• Correlation coefficient r (e.g., r = 0.85): strong positive linear relationship; check scatter plot for nonlinearity or influential points.
• Regression equation (e.g., y = a + bx): use for prediction within the observed range; report goodness-of-fit (r^2) and caution about extrapolation.

Common cautions:

• Do not infer causation from correlation without controlled experiments or additional evidence.
• Watch for hidden averages (Simpson's paradox) where aggregated trends reverse when data are disaggregated.
• Report sample size and measures of uncertainty where possible; small samples can give unstable estimates.