Finding the Median in an Even-Sized Data Set

Why an Even-Sized Data Set Has Two Middle Values

A median marks the center of an ordered data set. When the number of observations is odd, one observation occupies the central position. When the number is even, the center falls between two positions instead. The standard convention is to average the values in those two positions. As a result, the median is a measure of center, but it need not be one of the values recorded.

The key is to separate two tasks: first, put the observations in order; then, use the sample size to find the central positions. Averaging the wrong pair, or averaging before sorting, can produce an incorrect result.

The sample size \(n\) is even, so \(n/2\) is a whole number. The central observations are therefore the values immediately on either side of the midpoint between positions \(n/2\) and \(n/2+1\). This rule applies whether the two values are distinct or equal.

Finding the Central Positions

It can help to write the sample size as \(n=2m\), where \(m\) is a whole number. The two middle positions are then \(m\) and \(m+1\). There are \(m-1\) observations before the middle pair, and \(m-1\) after it. If the data are divided into two groups of \(m\) observations each, the first group ends with the lower middle value and the second group begins with the upper middle value.

For example, if \(n=10\), the middle positions are 5 and 6. Four observations precede this pair, in positions 1 through 4, and four follow it, in positions 7 through 10. If the ordered data are split into positions 1 through 5 and positions 6 through 10, each group contains five observations: the lower middle value is the last value in the first group, and the upper middle value is the first value in the second group.

Why Averaging the Middle Pair Is a Meaningful Convention

Suppose the ordered observations are \(x_{(1)}, \ldots, x_{(2m)}\), and let the two central values be \(a=x_{(m)}\) and \(b=x_{(m+1)}\). Since the observations are ordered, \(a\leq b\). The median is the midpoint \((a+b)/2\), which lies between \(a\) and \(b\), including the endpoints when they are equal.

This midpoint has a useful central property. The first \(m\) observations are all at or below \(a\), and \(a\) is at or below the median. Thus at least \(m\) observations are at or below the median. The last \(m\) observations are all at or above \(b\), and \(b\) is at or above the median. Thus at least \(m\) observations are at or above the median. Since \(n=2m\), the median has at least half the observations on each side when values equal to it are included.

This argument also explains why the median does not have to be an observed value. If \(a<b\), their midpoint may be a value strictly between them. The data themselves have no observation in the gap, but the median still identifies the center between the two middle observations. If \(a=b\), the average equals that shared value, so the median is observed.

Work From the Original Data Carefully

In many problems, the observations are not supplied in order. Sorting is essential because the middle positions refer to rank, not to the order in which values happen to be written. Repeated values remain separate observations: if a value occurs three times, it occupies three positions in the ordered list.

A repeated middle value is also possible. For example, in the ordered data 5, 7, 7, 7, 10, 12, the middle positions are 3 and 4, and both values are 7. The median is \((7+7)/2=7\), which is an observed value. Repetition does not change the procedure; it simply affects which values occupy the central positions.

Common Mistakes and Checks

• Choosing a middle value without ordering the data. The central positions are defined by rank. Sort the observations before identifying them.
• Averaging the observations in the middle of the written list. Their locations in the original list do not matter; use positions \(n/2\) and \(n/2+1\) in the ordered list.
• Using only one of the two central values. With an even sample size, the median is the average of both central values, not just the lower or upper one.
• Dropping repeated observations. Each recorded value counts, even if several observations have the same value.
• Assuming the median must be observed. When the central values differ, their average may fall between recorded values. Report it as the median, with units.
• Miscounting what lies before and after the middle pair. For ten observations, positions 5 and 6 are the middle pair; only positions 1–4 precede it and positions 7–10 follow it.

A quick check is to confirm that the chosen values are in the two central positions and that the calculated median lies between them. If the values are 10 and 12 centimeters, for instance, their average must be 11 centimeters, not a value outside that range.