Professional Documents
Culture Documents
Skip to content
Skip to primary sidebar
Home
About Me
Contact Me
Statistics By Jim
Basics
Hypothesis Testing
Regression
ANOVA
Fun
Glossary
Blog
Recommendations
A measure of central tendency is a summary statistic that represents the center point or
typical value of a dataset. These measures indicate where most values in a distribution
fall and are also referred to as the central location of a distribution. You can think of it as
the tendency of data to cluster around a middle value. In statistics, the three most
common measures of central tendency are the mean, median, and mode. Each of these
measures calculates the location of the central point using a different method.
Choosing the best measure of central tendency depends on the type of data you have.
In this post, I explore these measures of central tendency, show you how to calculate
them, and how to determine which one is best for your data.
The three distributions below represent different data conditions. In each distribution,
look for the region where the most common values fall. Even though the shapes and
type of data are different, you can find that central location. That’s the area in the
distribution where the most common values are located.
As the graphs highlight, you can see where most values tend to occur. That’s the
concept. Measures of central tendency represent this idea with a value. Coming up,
you’ll learn that as the distribution and kind of data changes, so does the best measure
of central tendency. Consequently, you need to know the type of data you have, and
graph it, before choosing a measure of central tendency!
Mean
The mean is the arithmetic average, and it is probably the measure of central tendency
that you are most familiar. Calculating the mean is very simple. You just add up all of
the values and divide by the number of observations in your dataset.
The calculation of the mean incorporates all values in the data. If you change any value,
the mean changes. However, the mean doesn’t always locate the center of the data
accurately. Observe the graphs below where I display the mean in the distributions.
Median
The median is the middle value. It is the value that splits the dataset in half. To find the
median, order your data from smallest to largest, and then find the data point that has
an equal amount of values above it and below it. The method for locating the median
varies slightly depending on whether your dataset has an even or odd number of
values. I’ll show you how to find the median for both cases. In the examples below, I
use whole numbers for simplicity, but you can have decimal places.
In the dataset with the odd number of observations, notice how the number 12 has six
values above it and six below it. Therefore, 12 is the median of this dataset.
When there is an even number of values, you count in to the two innermost values and
then take the average. The average of 27 and 29 is 28. Consequently, 28 is the median
of this dataset.
Outliers and skewed data have a smaller effect on the median. To understand why,
imagine we have the Median dataset below and find that the median is 46. However, we
discover data entry errors and need to change four values, which are shaded in the
Median Fixed dataset. We’ll make them all significantly higher so that we now have a
skewed distribution with large outliers.
As you can see, the median doesn’t change at all. It is still 46. Unlike the mean, the
median value doesn’t depend on all the values in the dataset. Consequently, when
some of the values are more extreme, the effect on the median is smaller. Of course,
with other types of changes, the median can change. When you have a skewed
distribution, the median is a better measure of central tendency than the mean.
Now, let’s test the median on the symmetrical and skewed distributions to see how it
performs, and I’ll include the mean so we can make comparisons.
In a symmetric distribution, the mean and median both find the center accurately. They
are approximately equal.
In a skewed distribution, the outliers in the tail pull the mean away from the center
towards the longer tail. For this example, the mean and median differ by over 9000, and
the median better represents the central tendency for the distribution.
These data are based on the U.S. household income for 2006. Income is the classic
example of when to use the median because it tends to be skewed. The median
indicates that half of all incomes fall below 27581, and half are above it. For these data,
the mean overestimates where most household incomes fall.
When to use the median: Skewed distribution, Continuous data, Ordinal data
Mode
The mode is the value that occurs the most frequently in your data set. On a bar chart,
the mode is the highest bar. If the data have multiple values that are tied for occurring
the most frequently, you have a multimodal distribution. If no value repeats, the data do
not have a mode.
In the dataset below, the value 5 occurs most frequently, which makes it the mode.
These data might represent a 5-point Likert scale.
Typically, you use the mode with categorical, ordinal, and discrete data. In fact, the
mode is the only measure of central tendency that you can use with categorical data—
such as the most preferred flavor of ice cream. However, with categorical data, there
isn’t a central value because you can’t order the groups. With ordinal and discrete data,
the mode can be a value that is not in the center. Again, the mode represents the most
common value.
In the graph of service quality, Very Satisfied is the mode of this distribution because it
is the most common value in the data. Notice how it is at the extreme end of the
distribution. I’m sure the service providers are pleased with these results!
In the continuous data below, no values repeat, which means there is no mode. With
continuous data, it is unlikely that two or more values will be exactly equal because
there are an infinite number of values between any two values.
When you are working with the raw continuous data, don’t be surprised if there is no
mode. However, you can find the mode for continuous data by locating the maximum
value on a probability distribution plot. If you can identify a probability distribution that
fits your data, find the peak value and use it as the mode.
The probability distribution plot displays a lognormal distribution that has a mode of
16700. This distribution corresponds to the U.S. household income example in the
median section.
When to use the mode: Categorical data, Ordinal data, Count data, Probability
Distributions
When you have ordinal data, the median or mode is usually the best choice. For
categorical data, you have to use the mode.
In cases where you are deciding between the mean and median as the better measure
of central tendency, you are also determining which types of statistical hypothesis
testsare appropriate for your data—if that is your ultimate goal. I have written an article
that discusses when to use parametric (mean) and nonparametric (median) hypothesis
testsalong with the advantages and disadvantages of each type.
Share this:
o
o inShare
o
o
o
Reply
2. John says
Reply
Reply
Hi Jim,
1) It seems like a good tie-in to this article would be one that describes box plots and
how to understand the information that they provide. Is there an article that you’ve
written on box plots that could be linked to this?
2) Unless I’m mistaken, the central tendency of a distribution and variability around that
central value tie into the concepts of accuracy and precision. Any chance that you could
speak to those concepts in a future article?
Reply
Hi Chuck,
Thank you very much! Those are both great ideas too.
I definitely plan to write a more comprehensive post about how the various aspects of
distributions work together–which would be a natural place to show box plots. I haven’t
written that yet but it is on my list of things to write about this spring.
As for accuracy and precision, we definitely have very specific definitions for those terms in
statistics. While in everyday English they are often considered synonyms, in statistics they’re
very different. And, you’re correct, they do tie into those two concepts. These terms often
come up in measurement system analysis.
If you measure parts repeatedly and the average or central tendency of the measurements
are unbiased (on target on average), you have an accurate measurement system. However,
if the measurements are biased (systematically too high or too low), your measurement
system is inaccurate.
If you measure the same part multiple times and the variability between measurements is
low, your measurement system is precise. However, if the measurements vary quite a bit,
your system is imprecise.
You can have any combination of accuracy and precision. Accurate and precise. Accurate
but not precise. Not accurate but precise. Neither accurate nor precise.
Reply
4. Chuck Wynn says
Thanks for that response Jim. I have one more quick question. I would think that the
mode for continuous data would be important when it comes to distributions that have
two (bimodal) or more (multi-modal) peaks. In these cases, where one has more than
one center of tendency, it would seem to me that the mode measure of central tendency
becomes the more important piece of information than either the mean or median.
Would this be accurate? Or is the answer, “It depends”? 🙂
Reply
Hi Chuck! Apologies for the delay in getting back to you. I’ve been on vacation!
I agree with what you say about multimodal continuous distributions. In fact, if you have a
multimodal distribution, it’s often crucial that you make that determination. Suppose that you
use a histogram to display the distribution of body heights. You notice that there are two
peaks. There are at least three important issues here.
1) If you are trying to identify the best probability distribution for your data, you won’t
succeed!
2) You also know that there is something else of interest for you to learn about your data.
For our example, the two peaks might indicate separate measures of central tendencies for
males and females. You can then better understand your data and how to analyze it.
3) As you mention, the mean and median are less meaningful for the single multimodal
distribution. You’ll probably want to identify the subpopulations (if they exist) and change
your analysis.
Graphing is always important for understanding your data. In this case, you do want to know
about multimodal distribution because it affects how you interpret the measure of central
tendency and could very well change how you analyze your data. It can actually point you to
understanding something new about your data. In the example about the heights, we
learned that males and females each have their own distribution. Gender is a relevant
variable in our analysis. That’s a fairly obvious example. However, in other cases, it might
lead you to something that you didn’t already consider. It’s a bit like being a detective and
looking for clues!
Reply
5. photonsquared says
Jim, how do you handle data spread when not using the mean?
Reply
Hi, you must be psychic! I’m writing a post about different measures of variability right now! If
you’re not using the mean because your data are skewed, I find that using the median for
the central tendency and interquartile range (IQR) for the variability goes together nicely.
The median splits that data in half and the IQR tells you where the middle half of the data
fall. The wider the IQR, the greater the spread the data spread. You can also use percentiles
to determine the spread for other proportions. For example, 95% of the data fall between the
2.5th and 97.5th percentiles.
Reply
6. Manas says
Reply
Reply
7. christian says
Reply
Hi Christian,
In many distributions, there are values that are more likely and less likely to occur. A
measure of central tendency identifies where values are more likely to occur–or where they
*tend* to occur. Hence, “tendency.”
Central is more applicable to the mean and median. Both of these measures identify a
central point in the distribution. This central point is where the values are more likely to
occur.
As we saw in the post with categorical data, there is no central value. Consequently, central
doesn’t really apply for the mode. But, we still use the terminology.
Reply
8. Sakshi Sharma says
Reply
Reply
9. learning262 says
Jim , It has been a while since my last stat class and i needed a refresh on those basic
unfortunately a mooc the cost hundereds of dollars could not help . Your articles helped
my greatly and i love the intuitive approach . I am going to go through all your articles ,
Please do keep writing more
Reply
Hllo sir ..
I am ur biggest fan .
Gve ur posts on Distributions of Probability
…hope I wll get ths soon
Reply
Hi Khursheed, I’ve already written that post. You can find it here: Understanding Probability
Distributions.
Reply
11. Carol says
Hi Jim,
thanks for your blog. I found this searching for information, and so far it’s been the
easiest to read and understand! I’m attempting to do a course on using and interpreting
data in schools and I get so confused with it all! When someone asks about the
relationship between the mean and the median, what are they asking for?
Reply
Hi Carol,
Thanks! I’m glad to hear it’s been helpful! I strive to make my blog as easy to understand as
possible.
That is a bit of a vague question, but I hope the context in which it’s being asked helps.
What they might be asking for is a description of how the mean and median tend to be
approximate equal for symmetric distributions. As the distribution becomes more skewed,
the difference between the mean and median increases, with the mean being pulled towards
the long tail.
Leave a Reply
Primary Sidebar
Meet Jim
I’ll help you intuitively understand statistics by focusing on concepts and using plain
English so you can concentrate on understanding your results.
Read More…
Follow Me
RSS Feed
Twitter
Popular
Latest
How To Interpret R-squared in Regression Analysis
How to Interpret P-values and Coefficients in Regression Analysis
Understanding Interaction Effects in Statistics
How to Interpret the F-test of Overall Significance in Regression Analysis
The Importance of Statistics
Choosing the Correct Type of Regression Analysis
Standard Error of the Regression vs. R-squared
Copyright © 2018 · Jim Frost