Histograms [AL]
Histograms as a Method of Data Presentation
A histogram is a graphical display used to show the frequency distribution of a continuous variable. It is one of the most important tools for understanding the shape of a dataset — revealing whether data are approximately normally distributed, skewed, bimodal, or otherwise non-standard. Understanding how to construct and interpret histograms correctly is an essential research skill, particularly because they are frequently confused with bar charts.
Structure of a Histogram
In a histogram, the x-axis represents the values of the variable being measured, divided into equal-width class intervals (also called bins). The y-axis represents the frequency — the number of observations falling within each interval. The height of each bar is proportional to the number of data points in that interval.
The defining structural feature of a histogram is that there are no gaps between adjacent bars. This reflects the continuous nature of the variable — any value on the measurement scale could in principle occur, and adjacent bars represent adjacent, unbroken intervals on this scale. The absence of gaps is not merely aesthetic; it communicates that the underlying variable is continuous.
How Histograms Differ from Bar Charts
| Histogram | Bar chart | |
|---|---|---|
| Data type | Continuous (interval/ratio) | Discrete categories or conditions |
| Gaps between bars | No gaps | Gaps present |
| x-axis | Measurement scale (e.g. score range) | Named categories or conditions |
| Purpose | Show frequency distribution shape | Compare means or frequencies across groups |
| y-axis | Frequency (count in each interval) | Mean, frequency, or percentage |
What Histograms Reveal
The shape of a histogram provides important information about the distribution of data:
- Normal distribution: a symmetrical, bell-shaped histogram — most scores cluster around the centre, with frequencies decreasing symmetrically towards both tails.
- Positive skew: a histogram with a long tail extending to the right — most scores are low, with a few high outliers pulling the tail rightward.
- Negative skew: a long tail extending to the left — most scores are high, with a few low outliers.
- Bimodal distribution: two distinct peaks — suggesting two subgroups within the data.
- Outliers: isolated bars far from the main distribution — visible as isolated bars with a gap (even within a histogram, individual extreme values may appear isolated).
This distributional information is critical for choosing appropriate statistical procedures — parametric tests (t-test, Pearson's r) assume approximately normal distributions, which can be assessed visually from a histogram before running the test.
Choosing Class Interval Width
The width of the class intervals affects the appearance and informativeness of the histogram. Too few, wide intervals (e.g. grouping all scores into just 3 bins) produces a coarse, uninformative display. Too many, narrow intervals produce a jagged display where every score appears as its own bar. The optimal number of intervals is typically between 5 and 15, chosen to reveal the distribution's shape clearly without excessive noise.
Key Takeaways
- Histogram: graphical display of the frequency distribution of a continuous variable — shows how often scores fall in each interval.
- No gaps between bars: the defining feature — reflects the continuous nature of the x-axis variable. Gaps would imply discrete categories.
- Bar chart vs histogram: bar chart = discrete categories (gaps); histogram = continuous variable (no gaps). Mixing these up is a fundamental error.
- Histogram shapes: normal (symmetrical bell), positive skew (tail right), negative skew (tail left), bimodal (two peaks).
- Histograms allow visual assessment of normality — essential before applying parametric statistical tests that assume normal distributions.
- Class interval width: too few bins = coarse and uninformative; too many = jagged and noisy. Aim for 5–15 intervals that reveal distribution shape clearly.