# What Is Continuous Data? Definition and Examples

What is continuous data? It is data that can take any value within a range, including fractions and decimals, because it comes from measurement rather than counting. Height, weight, temperature and time are the classic examples. If you can keep dividing the scale and still get a meaningful value, you are looking at continuous data.

## Quick Answer

- Continuous data comes from measuring, so it can take any value in an interval, such as 172.4 cm or 36.65 °C.
- Discrete data comes from counting, so it takes separate values, usually whole numbers, such as 0, 3 or 12 siblings.
- The test is simple: between any two values, can a third meaningful value exist? If yes, the data is continuous.
- Continuous variables are usually summarized with a mean and a standard deviation, and displayed with a histogram or box plot.
- Precision is limited by your instrument, not by the variable itself. A scale that reads to 0.1 cm still measures a continuous quantity.

## What Continuous Data Means

In plain terms, continuous data is any measurement that can be split into smaller and smaller pieces without losing its meaning. A person's height is not really 175 cm. It is 175.0 cm, or 175.04 cm, or 175.038 cm, depending on how precise your ruler is. The underlying quantity has no gaps.

The precise statistical definition is narrower. A continuous random variable is one whose set of possible values is an interval of real numbers, and whose probability is described by a probability density function. For a continuous variable, the probability of any single exact value is zero. You can only talk about the probability that the value falls inside a range, such as the probability that a height is between 170 cm and 180 cm.

This is the key difference from a discrete variable, where each value carries its own probability. If you roll a die, the probability of a 4 is one sixth. If you measure a height, the probability of exactly 175.0000 cm is zero.

Many continuous variables in practice are modeled with the normal distribution, which is symmetric and bell shaped and described by a mean $\mu$ and a standard deviation $\sigma$ [1]. That model is a convenience, not a requirement. Plenty of continuous data, such as income or waiting times, is skewed and needs a different distribution.

## How It Works

The mechanism behind continuous data is measurement. A measuring instrument maps a real quantity onto a number line, and the resolution of that instrument decides how many decimals you record.

The normal probability density function is the most common model for continuous measurements:

$$ f(x) = \frac{1}{\sigma \sqrt{2\pi}} e^{-\frac{1}{2}\left(\frac{x-\mu}{\sigma}\right)^2} $$

Each symbol has a job:

- $x$ is the value of the measurement, such as a height in centimeters.
- $\mu$ (mu) is the population mean, the center of the distribution.
- $\sigma$ (sigma) is the population standard deviation, which controls how spread out the curve is.
- $\sigma^2$ is the variance, the square of the standard deviation.
- $f(x)$ is the density at $x$, not a probability. To get a probability you integrate $f(x)$ over an interval.

The shorthand $X \sim N(\mu, \sigma^2)$ says that the variable $X$ is normally distributed with those two parameters [1].

For a sample, you estimate the center and spread with the sample mean and the sample standard deviation:

$$ \bar{x} = \frac{1}{n}\sum_{i=1}^{n} x_i \qquad s = \sqrt{\frac{\sum_{i=1}^{n}(x_i - \bar{x})^2}{n-1}} $$

Here $n$ is the number of observations, $x_i$ is the $i$-th measurement, and $\bar{x}$ is the sample mean. The $n-1$ in the denominator is the degrees-of-freedom correction that makes $s^2$ an unbiased estimate of the population variance.

## Worked Example

The dataset is a small class health survey with 15 measurements: 8 heights in centimeters, 4 sibling counts, and 3 body temperatures in Celsius.

| id | value | variable |
|---|---|---|
| H01 | 172.4 | height_cm |
| H02 | 168.1 | height_cm |
| H03 | 181.7 | height_cm |
| H04 | 175.0 | height_cm |
| H05 | 169.3 | height_cm |
| H06 | 178.6 | height_cm |
| H07 | 174.2 | height_cm |
| H08 | 183.5 | height_cm |
| C01 | 12 | siblings |
| C02 | 3 | siblings |
| C03 | 0 | siblings |
| C04 | 5 | siblings |
| T01 | 36.6 | temp_c |
| T02 | 37.1 | temp_c |
| T03 | 36.9 | temp_c |

Step 1 is classification. Height and temperature are continuous because they come from measurement and can take decimal values. Sibling count is discrete because it comes from counting and can only be a whole number. So the continuous variables are height_cm and temp_c, and the discrete variable is siblings.

Step 2 is to count the continuous height values: $n = 8$.

Step 3 is the sum of the heights: 1402.8 cm.

Step 4 is the mean height:

$$ \bar{x} = \frac{1402.8}{8} = 175.3500 \text{ cm} $$

Step 5 is the squared deviation of each height from the mean:

- $(172.4 - 175.35)^2 = 8.7025$
- $(168.1 - 175.35)^2 = 52.5625$
- $(181.7 - 175.35)^2 = 40.3225$
- $(175.0 - 175.35)^2 = 0.1225$
- $(169.3 - 175.35)^2 = 36.6025$
- $(178.6 - 175.35)^2 = 10.5625$
- $(174.2 - 175.35)^2 = 1.3225$
- $(183.5 - 175.35)^2 = 66.4225$

Step 6 is the sum of those squared deviations: $SS = 216.6200$.

Step 7 is the sample variance, using $n-1$ degrees of freedom:

$$ s^2 = \frac{216.6200}{8 - 1} = 30.9457 $$

Step 8 is the sample standard deviation:

$$ s = \sqrt{30.9457} = 5.5629 \text{ cm} $$

The same numbers come out of a short Python script:

```python
import statistics
heights = [172.4, 168.1, 181.7, 175.0, 169.3, 178.6, 174.2, 183.5]
mean_h = statistics.fmean(heights)
sd_h = statistics.stdev(heights)  # n-1, matches Excel STDEV.S
print(round(mean_h, 4), round(sd_h, 4))  # 175.35 5.5629
```

Output:

```text
175.35 5.5629
```

The mean height is 175.35 cm and the sample standard deviation is 5.5629 cm. The standard deviation is in the same units as the data, so you can read it directly: a typical height sits about 5.6 cm away from the mean.

## How to Interpret It

The mean tells you where the center of the measurements sits. The standard deviation tells you how tightly the values cluster around that center. For the heights above, a mean of 175.35 cm with a standard deviation of 5.56 cm puts one standard deviation either side of the mean at roughly 170 cm to 181 cm, and 4 of the 8 heights fall in that band.

Always report the units. A standard deviation of 5.5629 means nothing until you say it is 5.5629 cm. Units are part of the result, not decoration.

Look at the shape before you trust the mean. Plot the values and check whether they are roughly symmetric. If the distribution is skewed, the mean gets pulled toward the long tail and the median describes the typical value better. The normal model only fits when the data is roughly symmetric and bell shaped [1].

Check the order of the observations too. If you collected the data over time, plot the values in sequence. A run sequence plot can reveal that the location or spread is not constant over time, which makes distributional inferences unreliable [2]. A drifting mean is a process problem, not a distribution problem.

## When to Use It (and when not to)

Use continuous methods when the variable is genuinely measured on a continuous scale and you want to describe its center, spread or distribution. Means, standard deviations, histograms, box plots, t-tests and linear regression all assume you are working with continuous measurements.

Use them when the sample is small but the data is continuous. Methods for continuous data from small samples exist and are widely used in medical and laboratory work [3]. Small samples limit your power, not your choice of variable type.

Do not use continuous methods on counts. Sibling counts, number of visits, and number of defects are discrete. Treating them as continuous can produce nonsense, such as a predicted 2.4 children.

Do not use them on categories either. A variable coded 1 for male and 2 for female is numeric in your spreadsheet but not continuous, and the mean of that column is meaningless.

Be careful with rounded measurements. If everyone's height is recorded to the nearest centimeter, the data looks discrete in your file. The underlying variable is still continuous, and you should analyze it as such.

## Continuous Data vs Discrete Data

The closest related idea is discrete data, which comes from counting. The comparison below is the fastest way to tell them apart.

| Feature | Continuous data | Discrete data |
|---|---|---|
| Origin | Measurement | Counting |
| Possible values | Any value in an interval | Separate, usually whole values |
| Decimals | Meaningful, limited by instrument precision | Usually not meaningful |
| Example from the survey | Height 172.4 cm, temperature 36.6 °C | Siblings 0, 3, 5, 12 |
| Typical summary | Mean and standard deviation | Counts and proportions |
| Typical chart | Histogram, box plot, scatter plot | Bar chart |
| Probability of an exact value | Zero | Can be greater than zero |

The dividing line is whether a value can exist between two others. Between 172.4 cm and 172.5 cm there is 172.45 cm. Between 3 siblings and 4 siblings there is nothing.

## Common Mistakes

- **Calling any number with a decimal point continuous.** A price rounded to two decimals is still a measurement, but a rating of 4.0 on a 5-point scale is ordinal. Fix: ask how the value was produced, by measuring or by counting or ranking.
- **Treating counts as continuous.** Running a t-test on sibling counts ignores that the values are whole numbers. Fix: use methods built for counts, such as Poisson or negative binomial models.
- **Reporting a mean without a spread.** A mean of 175.35 cm alone hides whether the class is uniform or split into two groups. Fix: report the standard deviation or an interquartile range alongside the mean.
- **Ignoring the units.** A standard deviation of 5.5629 is ambiguous until you attach cm. Fix: carry units through every table and sentence.
- **Assuming normality because the variable is continuous.** Continuous does not mean bell shaped. Fix: plot the data and check the shape before choosing a parametric method [4].
- **Over-reading precision.** A scale that reads 175.35 cm does not mean the true height is known to that precision. Fix: round your reported results to a sensible number of digits.

## Limitations

Continuous data methods describe and compare groups, but they cannot tell you why a value is what it is. A mean height of 175.35 cm does not explain the growth, nutrition or genetics behind it. You still need a design and a causal argument.

The normal model is a convenience that fails on skewed or heavy-tailed data. Applying it blindly can give confidence intervals and p-values that are wrong. Small samples make this worse, because you cannot check the shape reliably and outliers have more influence on the mean and standard deviation [3]. When the distribution is clearly not normal, non-parametric methods are often the safer choice [4].

Finally, continuous data is only as good as the instrument. Measurement error, rounding and calibration drift all enter your numbers before you analyze them, and no statistical method can remove that error afterward.

## Frequently Asked Questions

### Is height continuous or discrete data?

Height is continuous. It can take any value in a range, and the only limit on how finely you record it is the precision of your measuring instrument. A recorded value of 175 cm is a rounded version of a continuous quantity, not a count.

### What is the difference between continuous and discrete data?

Continuous data comes from measuring and can take any value in an interval, including decimals. Discrete data comes from counting and takes separate values, usually whole numbers. Between two continuous values a third value always exists. Between two counts it does not.

### Can continuous data be counted?

You can count how many observations fall into a range, such as how many people are taller than 180 cm. That count is discrete. The underlying variable, height, stays continuous. The count is a summary of the continuous variable, not a replacement for it.

### What statistics do you use for continuous data?

The mean and standard deviation describe center and spread. A histogram or box plot shows the shape. For comparisons between groups, t-tests and analysis of variance are common when the data is roughly normal, and non-parametric tests are used when it is not [4].

### Does continuous data have to be normally distributed?

No. Continuous describes the set of possible values, not the shape of the distribution. Many continuous variables, such as income or waiting times, are skewed. Check the shape of your data before choosing a method that assumes normality [1].

## References

1. [6.5.1. What do we mean by "Normal" data?](https://www.itl.nist.gov/div898/handbook/pmc/section5/pmc51.htm)
2. [6.6.1.2. Graphical Representation of the Data](https://www.itl.nist.gov/div898/handbook/pmc/section6/pmc612.htm)
3. [Bland JM, Altman DG (2009). Analysis of continuous data from small samples. BMJ](https://doi.org/10.1136/bmj.a3166)
4. [Altman DG, Bland JM (2009). Parametric v non-parametric methods for data analysis. BMJ](https://doi.org/10.1136/bmj.a3167)

## Further Reading

- [NIST/SEMATECH e-Handbook of Statistical Methods](https://www.itl.nist.gov/div898/handbook/index.htm)
- [OpenStax. Introductory Statistics 2e](https://openstax.org/details/books/introductory-statistics-2e)

## Related Articles

- [What Is Discrete Data? Definition and Examples](/blog/data-analysis/what-is-discrete-data)
- [What Is Data? Definition, Meaning and Examples in Science](/blog/data-analysis/what-is-data-definition-meaning)
- [Quantitative Data Examples: Definition and Types](/blog/data-analysis/quantitative-data-examples)
- [Interval Data: Definition, Examples and When to Use It](/blog/data-analysis/interval-data-definition-examples)
- [What Is Structured Data? Definition and Examples](/blog/data-analysis/what-is-structured-data)
- [Discrete vs Continuous Variables: Key Differences](/blog/research-skills/discrete-vs-continuous-variables-key-differences)