In statistics, continuous describes a variable that can take on any value within a range, including decimals and fractions, with no gaps between possible values. Height, temperature, blood pressure, and time are all continuous because between any two measurements, there are infinitely many possible values. This is the core idea: you can always zoom in further, and there will always be more values in between.
Continuous vs. Discrete Variables
The easiest way to understand continuous is to contrast it with discrete. A discrete variable has countable, separate values. The number of children in a family is discrete: you can have 2 or 3, but not 2.7. A continuous variable has no such restriction. Someone’s weight could be 150 pounds, 150.3 pounds, 150.37 pounds, or 150.3718 pounds. Between any two values you pick, there’s an infinite number of values the measurement could theoretically land on.
This infinite divisibility is the defining feature. You might think you can list all possible weights between 150 and 151 pounds by going in increments of 0.01. But you’d skip 150.001, 150.0011, and so on forever. There is no way to count the number of values a continuous variable can take on.
Common Examples of Continuous Variables
Continuous variables show up everywhere in research and daily life. Blood pressure, age, body mass index, respiratory function, and the size of a tumor or lesion are all continuous. So are speed, distance, income, and reaction time. The unifying thread is that these measurements exist on a smooth scale where fractional values are meaningful. Saying someone is 34.6 years old or has a systolic blood pressure of 127.4 mm Hg makes sense, even if we rarely bother recording that level of detail.
This brings up an important nuance: continuous data often looks discrete once you record it. A thermometer might only read to the nearest tenth of a degree, and ages are typically rounded to whole years. The underlying variable is still continuous because the limitation is in the measuring tool, not in reality. Temperature doesn’t jump from 98.6 to 98.7 with nothing in between. Rounding is simply a practical shorthand to avoid false precision, where extra decimal places would imply accuracy the instrument doesn’t actually have.
Why a Single Point Has Zero Probability
One of the more counterintuitive facts about continuous variables is that the probability of landing on any single exact value is technically zero. If someone’s height can be any value between, say, 4 and 7 feet, the chance of being exactly 5.6234871… feet (to infinite decimal places) is vanishingly small. Instead, probability for continuous variables is always calculated over a range: the probability of being between 5.5 and 5.7 feet, for example.
This is why continuous variables use a probability density function (often called a PDF) rather than listing probabilities for individual values the way discrete variables do. The PDF is a smooth curve, and the probability of falling within any range equals the area under that curve between two points. Finding that area requires integral calculus, though in practice, software and statistical tables handle the math. The familiar bell curve of a normal distribution is the most recognized example of a PDF.
Measurement Scales for Continuous Data
Continuous variables fall into two measurement scales, and the distinction matters for how you can analyze them. Interval scale data has equal spacing between values but no true zero point. Temperature in Celsius is the classic example: the difference between 10°C and 20°C is the same as between 20°C and 30°C, but 0°C doesn’t mean “no temperature.” You can add and subtract meaningfully, but ratios don’t work (40°C is not “twice as hot” as 20°C in any physical sense).
Ratio scale data has both equal spacing and a true zero that represents the complete absence of the thing being measured. Weight, height, distance, and time all qualify. Zero kilograms means no mass. This true zero makes ratios meaningful: 10 kg is genuinely twice as heavy as 5 kg. Most continuous variables you’ll encounter in science and medicine are ratio scale.
How Continuous Data Is Visualized
The tools for visualizing continuous data differ from those used for categories or counts. Three methods form the foundation of exploratory analysis.
Histograms divide the range of values into bins and count how many observations fall into each one, giving you a picture of the distribution’s shape. Bin width matters more than most people realize: too few bins and you smooth over important patterns, too many and the chart becomes a jagged mess of random noise rather than meaningful signal. It’s worth experimenting with different bin widths rather than relying on whatever default your software chooses.
Box plots compress the distribution into a compact summary showing the median, the middle 50% of values (the interquartile range), and potential outliers. They’re especially useful for comparing a continuous variable across groups. Scatter plots display the relationship between two continuous variables, with each observation as a point on a two-dimensional plane.
Statistical Tests for Continuous Variables
The type of variable you’re working with determines which statistical tests are appropriate. Continuous data opens the door to a set of powerful parametric tests, provided the data is roughly normally distributed (following a bell curve shape).
To compare the averages of two groups, you’d use a t-test (unpaired for independent groups, paired for before-and-after measurements on the same subjects). For three or more groups, analysis of variance (ANOVA) replaces the t-test. To measure the strength of a linear relationship between two continuous variables, Pearson’s correlation coefficient is the standard choice.
When continuous data is skewed or doesn’t follow a normal distribution, nonparametric alternatives step in. The Mann-Whitney test replaces the unpaired t-test, and the Kruskal-Wallis test replaces ANOVA. These tests work with ranked values rather than raw numbers, making them less sensitive to extreme values.
Handling Outliers in Continuous Data
Because continuous variables can theoretically take on any value in a range, extreme observations (outliers) are a constant practical concern. There’s no universal agreement on how to define them. Some researchers flag any value more than two or three standard deviations from the mean. Others use the interquartile range method, where outliers are values falling more than 1.5 times the interquartile range above or below the middle 50%. Some simply identify them visually on a plot.
What to do with outliers is equally debated. A survey of research faculty found that only about 47% reported using all data including outliers when describing a continuous variable. Around 13% switched to reporting the median and interquartile range instead of the mean, which is less distorted by extreme values. About 11% ran a formal statistical test for outliers before deciding. The key takeaway is that removing outliers without a clear, pre-specified reason can inflate false positive rates, making results look more significant than they actually are.
Why It Matters to Get the Classification Right
Treating a continuous variable as though it were categorical, a process called dichotomizing, throws away information. Researchers sometimes do this for convenience, splitting blood pressure into “high” and “normal” or tumor size into “large” and “small.” But this collapses a rich, nuanced measurement into two buckets. A person with a systolic blood pressure of 139 gets classified differently from someone at 141, despite a trivial real difference, while someone at 141 and someone at 180 get lumped together.
Research in neuroradiology has shown that categorizing continuous variables this way can change study conclusions entirely, because the cutoff points are often chosen based on the data itself rather than any biological rationale. Keeping continuous data continuous preserves statistical power and produces more reliable results.