How to Find the Median of Discrete and Continuous Variables

The same definition of a median works for discrete, continuous, and mixed random variables. The calculation differs because a discrete distribution has jumps in its cumulative probability, while a continuous distribution usually accumulates probability smoothly.

The most reliable method is to use the cumulative distribution function (CDF). It also handles cases where a distribution has more than one median.

The median definition

The CDF of a random variable $X$ is $F(x)=P(X\le x)$. A number $m$ is a median if

$$P(X\le m)\ge \frac12 \quad\text{and}\quad P(X\ge m)\ge \frac12.$$

Equivalently, using the left-hand limit $F(m-)=P(X<m)$,

$$F(m-)\le \frac12\le F(m).$$

This form matters when $X$ can equal $m$ with positive probability. A common convention is to report the lower 0.5-quantile, $\inf\{x:F(x)\ge 1/2\}$. When the median is not unique, this convention picks one particular median.

Finding a median for a discrete variable

For a discrete variable, list the possible values in increasing order and add their probabilities cumulatively. The first value where the cumulative probability reaches or exceeds $0.5$ is the lower 0.5-quantile. Check the median conditions if you need to describe all medians rather than report one.

For example, suppose $X$ has this distribution:

$x$	$P(X=x)$	Cumulative probability $F(x)$
$-1$	$0.20$	$0.20$
$0$	$0.35$	$0.55$
$2$	$0.45$	$1.00$

The cumulative probability first reaches $0.5$ at $x=0$, so the lower 0.5-quantile is $0$. It is a median because $P(X\le0)=0.55$ and $P(X\ge0)=0.80$, both at least one half.

A discrete distribution can have an interval of medians. For example, if $P(X=0)=P(X=1)=0.5$, every $m$ between 0 and 1 satisfies both median conditions. The lower-quantile convention reports 0, but it is not the only median under the two-sided definition.

Finding a median for a continuous variable

For a continuous variable with density $f$, calculate the CDF by integrating the density. A median usually solves $F(m)=0.5$, or equivalently, the area under the density to the left of $m$ is one half.

For an exponential random variable with rate $2$, the CDF for $x\ge0$ is

$$F(x)=1-e^{-2x}.$$

Set this equal to one half and solve:

$$\begin{aligned} 1-e^{-2m}&=\frac12,\\ e^{-2m}&=\frac12,\\ m&=\frac{\ln 2}{2}\approx0.3466. \end{aligned}$$

Thus the median is about $0.3466$ in the same units as $X$. If the CDF is flat across a range at probability $0.5$, every point in that range is a median; strict increase is not required for a continuous variable.

What if the variable is mixed?

A mixed variable has both point probabilities and a continuous part. Use the same CDF rule: include the point masses as jumps, and add the continuous probability accumulated between them.

Suppose $P(X=0)=0.4$, and with probability $0.6$, $X$ is uniform on $[0,2]$. For $0\le x\le2$,

$$F(x)=0.4+0.6\frac{x}{2}=0.4+0.3x.$$

The median is where this reaches $0.5$:

$$0.4+0.3m=0.5 \quad\Longrightarrow\quad m=\frac13.$$

At this value, half the total probability is at or below $m$ and half is at or above it. The atom at zero creates a jump in the CDF, but it does not change the method.

Common mistakes

• Using the density value as the median. A median is about cumulative probability, not the value of the density $f(m)$. Integrate the density to obtain $F(m)$.
• Treating a discrete jump like a smooth crossing. For discrete variables, $F$ can jump from below $0.5$ to above it. The median may be the value at that jump; you do not need an outcome with cumulative probability exactly $0.5$.
• Assuming there is always one median. Flat regions or probability split evenly across separated points can produce multiple medians. State which convention you use if you need one number.
• Confusing a population median with a sample convention. For an even-sized data set, the sample median is commonly defined as the average of the two middle observations. A random variable’s median is instead defined by its probability on either side, so that average need not be a possible value of the variable.