Descriptive Statistics Refresher

MFT Topics discussed in these review slides are in bold:

1. Measure of set operations
2. Conditional/joint probabilities
3. Counting rules
4. Measures of central tendency and dispersion
5. Distributions (including normal and binomial)
6. Sampling and estimation
7. Hypothesis testing
8. Correlation and regression
9. Time-series forecasting
10. Statistical concepts in quality control
“SOCS”
When you analyze a set of data, remember “SOCS”
Shape – what is the shape of the distribution?
Outliers – Are there any unusual data values in the distribution?
Center – How can I best describe the typical value or center of the
distribution?
Spread – How can I describe the variation or dispersion in the data?
Commonly observed shapes
Modality:

Skewness or symmetry:
Histogram
Common graph for quantitative data. The horizontal axis is a number
line broken into ranges and the vertical axis is the count or frequency.

Shape: skewed left

Outliers
An outlier is a data value that appears extreme relative to the rest of
the data.
Outliers can often be identified by examining an appropriate graph of
the data.
Some outliers are data entry or data collection errors that can be
corrected after they are identified. Other outliers are natural features
of the data.
Since outliers impact many calculations, you should inspect your data
for outliers near the beginning of your analysis.
Center or central tendency
Measures of central tendency the center, or middle, or typical value of
a distribution.
Common measure of center: mean and median

Mean: The sample mean, denoted as x̄, can be calculated as

where x1, x2, ..., xn represent the n observed values.

Mean
The population mean is also computed the same way but is denoted as
µ. It is often not possible to calculate µ since population data are rarely
available.
The sample mean is a sample statistic, and serves as a
point estimate of the population mean. This estimate may not be
perfect, but if the sample is good (representative of the population),
it is usually a pretty good estimate.
The sample mean is impacted by the presence of outliers. The mean is
pulled toward the side of the distribution containing outliers.
Median
The median is the value that splits the data in half when ordered in
ascending order.

If there are an even number of observations, then the median is the

average of the two values in the middle.

Since the median is the midpoint of the data, 50% of the values are
below it. Hence, it is also the 50th percentile.
Variation or dispersion
Dispersion refers to the degree of variation in the data; that is, the
numerical spread (or compactness) of the data. How spread out is the
data?
As variation in the data increases, all measures of variation take larger
values.
Range
The range is the simplest measure of variation. The range is the
difference between the maximum value and the minimum value in the
data set.
Range = max value – min value

The range is affected by outliers, and is often used only for very small
data sets.
Variance
The variance is roughly the average squared deviation from the mean.
Here is the formula for the sample variance:

The sample variance is a point estimate of the corresponding

population variance. The population variance has a slightly different
formula:
Standard deviation
The standard deviation is the square root of the variance. The unit of
measure of the standard deviation is the same as the data. This makes
the standard deviation more practical than the variance to use in
applications. The SD is interpreted as roughly the average distance
between observations in the data set and the mean of the data.
Formula:
Empirical rule (or 68-95-99.7 rule)
In a unimodal, symmetric distribution, about 68% of the values fall
within one standard deviation of the mean, about 95% of the values
fall within two standard deviations of the mean, and about 99.7% of
the values fall within three standard deviations of the mean.
Quartiles
Quartiles are special percentiles that divide a data set into four
sections, each containing 25% of the data set.
Q1 = first quartile = 25th percentile
Q2 = second quartile = 50th percentile
Q3 = third quartile = 75th percentile

Lowest Middle Middle Highest

25% of 25% of 25% of 25% of
data data data data
Q1 Q2 Q3
Inter-quartile range
The interquartile range (IQR), or the midspread is the difference
between the first and third quartiles, Q3 – Q1.

This includes only the middle 50% of the data and, therefore, is not
influenced by extreme values.
Boxplots
Boxplots (or box-and-whisker plots) are graphical displays built from
the five-number summary. The five-number summary consists of the
min, Q1, median, Q3, and max.
The box extends from Q1 to Q3. A line is drawn inside the box at the
value of the median. The “whiskers” extend to the values of the min
and max.
In addition, boxplots are often modified to incorporate outlier
detection rules based on distances beyond the quartiles and either
1.5×IQR, for potential outliers, or 3×IQR for probable outliers.
 possible probable
possible outlier
outliers
outlier
1.5 × IQR 3.0 × IQR

o
* * *
inner inner outer x
fence fence fence

median

Q1 Q3
Z-scores
A standardized value, commonly called a z-score, provides a relative
measure of the distance an observation is from the mean, which is
independent of the units of measurement.

Subtracting the mean from all data values centers the data set at 0.
Dividing all of the centered values by the standard deviation scales the
values to a new standard deviation of 1.
The process of standardizing data with z-scores in often called
“centering and scaling” the data.
Impact of outliers
Outliers pull the mean toward them.
Outliers inflate the value of the range, variance, and standard
deviation.
(Note: measures of variation always get larger when outliers are
present.)
Outliers also impact other statistics, such as the correlation,
coefficients of regression models, etc.
Some statistics are resistant to the effects of outliers, like the median
and IQR.
Identifying outliers
There are several common rules of thumb for identifying outliers.
1) Values above Q3 + 1.5×IQR or below Q1 – 1.5×IQR, which are called
the “inner fences,” are potential outliers.
2) Values above Q3 + 3×IQR or below Q1 – 3×IQR, which are called the
“outer fences,” are probable outliers/extreme values.
3) Values with z-scores above +3 or below –3 are potential outliers.
Choosing appropriate measures
The mean and standard deviation are the most popular measures of
center and variation. If the data is roughly symmetric in shape and
contains no obvious outliers, these measures are acceptable.

The median and IQR, which are both resistant to the impact of outliers,
should be strongly considered when the data contains outliers or is
strongly skewed in shape.
Categorical data
Categorical data is fundamentally different than quantitative/numeric
data.
Averages, standard deviations, and other summary statistics often
make no sense for categorical data.
Sample proportion
The sample proportion, denoted by p or , is the fraction of data that
have a certain characteristic or that belong to a certain category.
Proportions are key descriptive statistics for categorical data, such as
defects or errors in quality control applications or consumer
preferences in market research.
Frequency distribution
A frequency distribution displays the values of a categorical variable
and one or more measures derived from the count of how often each
category occurs in the data.
Pie chart
Pie charts show the whole group of cases as a circle sliced into pieces
with sizes proportional to the fraction of the whole in each category.
Bar chart
A bar chart displays the distribution of a categorical variable, showing
the counts for each category next to each other for easy comparison.
Contingency table
The frequencies of two categorical variables can be summarized and
displayed simultaneously using a contingency table (or
crosstabulation):
Other bar charts
More elaborate bar charts, such as
clustered or stacked bar charts can
be created from contingency tables:
Practice
A) Which is most likely true for the distribution of “percentage of time
actually spent taking notes in class,” which is displayed in the
histogram?
(a) mean > median
(b) mean ~ median
(c) mean < median
(d) impossible to tell
Practice
B) Which of these variables do you expect to be uniformly distributed?
(a) weights of adult females
(b) salaries of a random sample of people from North Carolina
(c) house prices
(d) birthdays of classmates (day of the month)
Practice
C) If someone's gross annual income has a z-score of +2.3, what can be
concluded?
oTheir income is 2.3 standard deviations below the mean income.
oTheir income is 2.3 standard deviations above the mean income.
oTheir income is 2.3 times the mean income.
oTheir income is 2.3 standard deviations above the median income.
Practice
D) A community college school board is negotiating a new contract with the college
faculty. The distribution of faculty salaries is skewed right by several faculty
members who make over $100,000 per year. If the school board wants to give the
community the impression that the faculty are already overpaid, should they
adver se the mean or median of the faculty salaries?
o The school board should use the mean to make their argument. The mean will be
higher than the median since it will be influenced by the few high salaries.
o The school board should use the median to make their argument. The median
will be lower than the mean since the mean is influenced by the few high salaries.
o The school board should use the mean to make their argument. The mean will be
lower than the median since the median is influenced by the few high salaries.
Practice
E) A company advertises a mean lifespan of 1000 hours for a particular type
of light bulb. If you were in charge of quality control at the factory, would
you prefer that the standard deviation of the lifespans for the light bulbs be 5
hours or 50 hours? Why?
o50 hours would be preferable since a larger standard deviation indicates a
longer average lifespan for the light bulbs.
o5 hours would be preferable since a smaller standard deviation indicates
more consistency.
o50 hours would be preferable since a larger standard deviation indicates
more consistency.
o5 hours would be preferable since a smaller standard deviation indicates a
longer average lifespan for the light bulbs.
Practice solution
A) Which is most likely true for the distribution of “percentage of time
actually spent taking notes in class,” which is displayed in the
histogram?
(c) mean < median median: 80%
mean: 76%
The distribution is skewed
to the left and the data
values in the left tail pull
the mean toward them, but
the median is unaffected.
Practice solution
B) Which of these variables do you expect to be uniformly distributed?
(d) birthdays of classmates (day of the month)

C) If someone's gross annual income has a z-score of +2.3, what can be

concluded?
oTheir income is 2.3 standard deviations above the mean income.
Practice solution
C) A community college school board is negotiating a new contract with
the college faculty. The distribution of faculty salaries is skewed right by
several faculty members who make over $100,000 per year. If the
school board wants to give the community the impression that the
faculty are already overpaid, should they adver se the mean or median
of the faculty salaries?
oThe school board should use the mean to make their argument. The
mean will be higher than the median since it will be influenced by
the few high salaries.
Practice solution
E) A company advertises a mean lifespan of 1000 hours for a particular
type of light bulb. If you were in charge of quality control at the
factory, would you prefer that the standard deviation of the lifespans
for the light bulbs be 5 hours or 50 hours? Why?
o5 hours would be preferable since a smaller standard deviation
indicates more consistency.