\[ \begin{align}\begin{aligned}\newcommand\blank{~\underline{\hspace{1.2cm}}~}\\% Bold symbols (vectors) \newcommand\bs[1]{\mathbf{#1}}\\% Differential \newcommand\dd[2][]{\mathrm{d}^{#1}{#2}} % use as \dd, \dd{x}, or \dd[2]{x}\\% Poor man's siunitx \newcommand\unit[1]{\mathrm{#1}} \newcommand\num[1]{#1} \newcommand\qty[2]{#1~\unit{#2}}\\\newcommand\per{/} \newcommand\squared{{}^2} \newcommand\cubed{{}^3} % % Scale \newcommand\milli{\unit{m}} \newcommand\centi{\unit{c}} \newcommand\kilo{\unit{k}} \newcommand\mega{\unit{M}} % % Percent \newcommand\percent{\unit{{\kern-4mu}\%}} % % Angle \newcommand\radian{\unit{rad}} \newcommand\degree{\unit{{\kern-4mu}^\circ}} % % Time \newcommand\second{\unit{s}} \newcommand\s{\second} \newcommand\minute{\unit{min}} \newcommand\hour{\unit{h}} % % Distance \newcommand\meter{\unit{m}} \newcommand\m{\meter} \newcommand\inch{\unit{in}} \newcommand\foot{\unit{ft}} % % Force \newcommand\newton{\unit{N}} \newcommand\kip{\unit{kip}} % kilopound in "freedom" units - edit made by Sri % % Mass \newcommand\gram{\unit{g}} \newcommand\g{\gram} \newcommand\kilogram{\unit{kg}} \newcommand\kg{\kilogram} \newcommand\grain{\unit{grain}} \newcommand\ounce{\unit{oz}} % % Temperature \newcommand\kelvin{\unit{K}} \newcommand\K{\kelvin} \newcommand\celsius{\unit{{}^\circ C}} \newcommand\C{\celsius} \newcommand\fahrenheit{\unit{{}^\circ F}} \newcommand\F{\fahrenheit} % % Area \newcommand\sqft{\unit{sq\,\foot}} % square foot % % Volume \newcommand\liter{\unit{L}} \newcommand\gallon{\unit{gal}} % % Frequency \newcommand\hertz{\unit{Hz}} \newcommand\rpm{\unit{rpm}} % % Voltage \newcommand\volt{\unit{V}} \newcommand\V{\volt} \newcommand\millivolt{\milli\volt} \newcommand\mV{\milli\volt} \newcommand\kilovolt{\kilo\volt} \newcommand\kV{\kilo\volt} % % Current \newcommand\ampere{\unit{A}} \newcommand\A{\ampere} \newcommand\milliampereA{\milli\ampere} \newcommand\mA{\milli\ampere} \newcommand\kiloampereA{\kilo\ampere} \newcommand\kA{\kilo\ampere} % % Resistance \newcommand\ohm{\Omega} \newcommand\milliohm{\milli\ohm} \newcommand\kiloohm{\kilo\ohm} % correct SI spelling \newcommand\kilohm{\kilo\ohm} % "American" spelling used in siunitx \newcommand\megaohm{\mega\ohm} % correct SI spelling \newcommand\megohm{\mega\ohm} % "American" spelling used in siunitx % % Capacitance \newcommand\farad{\unit{F}} \newcommand\F{\farad} \newcommand\microfarad{\micro\farad} \newcommand\muF{\micro\farad} % % Inductance \newcommand\henry{\unit{H}} \newcommand\H{\henry} \newcommand\millihenry{\milli\henry} \newcommand\mH{\milli\henry} % % Power \newcommand\watt{\unit{W}} \newcommand\W{\watt} \newcommand\milliwatt{\milli\watt} \newcommand\mW{\milli\watt} \newcommand\kilowatt{\kilo\watt} \newcommand\kW{\kilo\watt} % % Energy \newcommand\joule{\unit{J}} \newcommand\J{\joule} % % Composite units % % Torque \newcommand\ozin{\unit{\ounce}\,\unit{in}} \newcommand\newtonmeter{\unit{\newton\,\meter}} % % Pressure \newcommand\psf{\unit{psf}} % pounds per square foot \newcommand\pcf{\unit{pcf}} % pounds per cubic foot \newcommand\pascal{\unit{Pa}} \newcommand\Pa{\pascal} \newcommand\ksi{\unit{ksi}} % kilopound per square inch % NOTE: do NOT define \bar -- it is the built-in LaTeX accent used by \bar{x}, % \bar{y} (sample means). Redefining it made those render as the word "bar". \newcommand\barunit{\unit{bar}} \end{aligned}\end{align} \]

Nov 16, 2026 | 3427 words | 34 min read

4.1.1. PCMs: Data Analysis and Histograms in Excel#

These pre-class materials introduce statistical tools that engineers use to analyze and interpret datasets using Microsoft Excel.

In engineering, data is often collected through experiments, testing, or simulations. To understand what the data means, engineers must summarize the results, compare different groups, and identify patterns within the dataset.

In this module, you will learn how to compute summary statistics, apply conditional calculations, and visualize how data values are distributed.

Descriptive Statistics#

Descriptive statistics are used to summarize and describe the characteristics of a dataset.

Rather than examining every individual value, engineers use descriptive statistics to understand:

  • Typical values within a dataset

  • Variation between data points

  • Differences between groups

  • Trends or patterns in data

These summaries help engineers interpret results, compare systems, and make decisions based on data.

Measures of Central Tendency#

Measures of central tendency describe the typical value of a dataset. These concepts were introduced in PCM: Spreadsheet Foundations.

Statistic

Definition

Excel Function

Mean

The average value of a dataset.

=AVERAGE()

Median

The middle value when the data are ordered. If the number of data points is even, the median is the average of the two middle values.

=MEDIAN()

Mode

The most frequent value in the dataset.

=MODE.SNGL()
=MODE.MULT()

If the number of data points is even, the median is the average of the two middle values after the data are ordered.

Measures of Variation#

In addition to understanding the center of a dataset, engineers must also understand how spread out the data values are. In this course, we will use three key measures of variation: range, variance, and standard deviation.

Range#

Range is the difference between the largest and smallest values in a dataset. It provides a quick way to estimate how widely the data values are spread.

Mathematical Expression

Excel Expression

\(\text{Range} = \text{max} - \text{min}\)

=MAX(range) - MIN(range)

Range is often useful as a quick way to identify potential outliers or unusually large or small values in a dataset. However, because it only uses the minimum and maximum values, it does not describe how the rest of the data are distributed.

For this reason, range is typically used alongside other measures of variation such as variance and standard deviation.

Check for Understanding

Which Excel expression correctly calculates the range of the values in cells A1:A10?

A. =RANGE(A1:A10)
B. =STDEV.S(A1:A10)
C. =AVERAGE(A1:A10)
D. =MAX(A1:A10)-MIN(A1:A10)

Show Answer

Correct Answer: D

Excel does not have a built-in RANGE() function. To calculate the range, subtract the minimum value from the maximum value:

=MAX(A1:A10) - MIN(A1:A10)

Standard Deviation and Variance#

Standard deviation and variance describe how far data values spread from the mean.

Variance measures the average squared difference between each value and the mean. A larger variance indicates greater spread in the dataset, while a smaller variance indicates that the data values are closer to the mean.

Standard deviation is the square root of the variance, which means it measures the same spread but in the same units as the original data, making it easier to interpret.

Measure

Definition

Excel Function

Standard Deviation

Measures how far data values typically spread from the mean.

Population: =STDEV.P()
Sample: =STDEV.S()

Variance

Measures the average squared difference between each data value and the mean.

Population: =VAR.P()
Sample: =VAR.S()

A sample represents a subset of a larger population. In engineering analysis, we often work with sample data because measuring every possible value in a population is not practical.

For this course, you will use the sample functions =STDEV.S() and =VAR.S().

Check for Understanding

Two datasets have the same mean. Dataset A has a larger standard deviation than Dataset B.

What does this indicate?

A. Dataset A has larger values than Dataset B
B. Dataset A’s values are more spread out from the mean
C. Dataset A has more data points
D. Dataset A has a larger median

Show Answer

Correct Answer: B

A larger standard deviation indicates that the data values are more spread out from the mean.

Logical Functions#

Logical functions allow Excel to evaluate conditions and return results based on whether those conditions are true or false.

These functions are useful when engineers need to:

  • classify results

  • check whether a value meets a requirement

  • apply rules to a dataset

  • automate decision-making within calculations

Logical functions test whether a condition is satisfied. The result of the test is either TRUE or FALSE.

In Excel calculations, these logical values can also behave like numbers in some situations:

  • TRUE corresponds to 1

  • FALSE corresponds to 0

Practice

Download and open the following file to follow along with the video:

Relational (Comparison) Operators#

Logical functions use comparison operators to test conditions.

Operator

Meaning

<

less than

>

greater than

<=

less than or equal to

>=

greater than or equal to

=

equal to

<>

not equal to

These operators allow Excel to compare values between cells.

Example: A1 > 10 This condition evaluates to:

  • TRUE if the value in A1 is greater than 10

  • FALSE if it is not

IF Function#

The IF function evaluates a condition and returns one value if the condition is TRUE and another value if the condition is FALSE.

General structure: =IF(condition, value_if_true, value_if_false)

Example: =IF(B2<30,"Freshman","Upperclassman") This formula checks whether the value in cell B2 is less than 30.

  • If the condition is TRUE, Excel returns “Freshman”

  • If the condition is FALSE, Excel returns “Upperclassman”

AND and OR Functions#

More complex conditions can be created using AND and OR functions.

Function

Purpose

AND()

Checks if all conditions are TRUE

OR()

Checks if any condition is TRUE

Example: =AND(A2>50, B2>50)
Returns TRUE only if both conditions are satisfied.

Example: =OR(A2>50, B2>50)
Returns TRUE if either condition is satisfied.

Nested IF Functions#

A nested IF function places one IF statement inside another IF statement.

This allows Excel to evaluate multiple conditions sequentially.

Example: =IF(A2>89,"A",IF(A2>79,"B",IF(A2>69,"C",IF(A2>59,"D","F"))))
This formula assigns a letter grade based on the value in A2.

Check for Understanding

What will Excel return if the following condition is evaluated?

=OR(5>10, 8>3)

A. TRUE
B. FALSE
C. ERROR
D. 8

Show Answer

Correct Answer: A

The first condition is FALSE, but the second condition is TRUE. Because OR() returns TRUE when any condition is true, the result is TRUE.

Histograms#

Engineers often use software tools such as Excel to generate histograms and analyze distributions.

A histogram is a chart used to show how numerical data are distributed across a range of values.

Instead of displaying individual data points, histograms group values into intervals called bins. Each bin represents a range of values, and the height of the bar shows how many data points fall within that range.

Histograms help engineers understand how data values are distributed and whether they cluster within certain ranges.

Practice

Download and open the following file to follow along with the video:

Components of a Histogram#

Histogram showing tree circumference distribution by bins and frequency.

Fig. 4.1 A histogram showing tree circumference values grouped into bins along the x-axis, with frequency shown on the y-axis.#

A histogram has several key parts:

  • Bins - ranges of numerical values used to group the data

  • Frequency (count) - the number of values that fall within each bin

  • X-axis - shows the ranges of values (the bins)

  • Y-axis - shows the frequency, or number of observations

Each bar represents how many data values fall within a specific range.

Bins divide the number line into intervals of equal width, making it easier to see how frequently values occur across different parts of the dataset.

Check for Understanding

In a histogram, the height of each bin represents ______ and is plotted along the ______.

A. Frequency (count); x-axis
B. Frequency (count); y-axis
C. Duration; y-axis
D. Frequency (count); percentage

Show Answer

Correct Answer: B

The height of each bin represents the frequency (count) of data values and is plotted on the y-axis.

Interval Notation for Bins#

Histogram bins are often written using interval notation, which indicates whether the endpoints of a range are included.

Two symbols are commonly used:

  • Parentheses ( ) -> value is not included

  • Brackets [ ] -> value is included

Example: \((30,70]\) means the bin contains values greater than 30 and up to and including 70. Or mathematically, \(30 < x \leq 70\)

This notation helps avoid counting values in more than one bin.

Check for Understanding

In a histogram, how should the bin labeled [20, 45) be interpreted?

A. The interval from 20 to 45, including 20 and 45
B. The interval from 20 to 45, including 20 but not including 45
C. The interval from 20 to 45, not including 20 but including 45
D. The interval from 20 to 45, not including 20 or 45

Show Answer

Correct Answer: B

The notation [20, 45) means the bin includes values greater than or equal to 20 and less than 45.

Check for Understanding

Why do histogram bins often use interval notation such as (20, 40] or [40, 60)?

A. To make the histogram easier to draw
B. To make the bins the same width
C. To ensure each data value is counted in only one bin
D. To show that the data are continuous

Show Answer

Correct Answer: C

Interval notation helps ensure that each data value belongs to only one bin, preventing values at the boundaries from being counted twice.

Choosing the Number of Bins#

The number of bins used in a histogram affects how clearly the distribution can be interpreted.

  • Too few bins may hide important patterns in the data.

  • Too many bins may make the histogram appear noisy or difficult to interpret.

The goal is to choose bins that clearly show the overall distribution and spread of the dataset.

Calculating Number of Bins and Bin Size#

  1. Determine the number of bins needed using the formula: \(\text{Number of bins} = \sqrt{N}\), rounded to a wholenumber, where N is the sample size, or the total number of data points in the data set. Hint: In Excel, use =COUNT( ) to quickly find N.

  2. Determine the range of the data.

  3. Determine the number of items in each bin using the formula: \(\text{Bin size} = \text{range} / \text{number of bins}\). Note that bin size is often rounded to a whole number.

What Histograms Can Show#

Histograms help engineers analyze datasets in several ways:

  • Understanding the spread of the data

  • Determining whether values are clustered in certain ranges

  • Identifying possible outliers

  • Comparing the distributions of different datasets

  • Observing whether a dataset follows a normal (bell-shaped) distribution

Histograms vs Bar Charts#

Histograms and bar charts may look similar, but they represent different types of data.

Histogram

Bar Chart

Used for continuous numerical data

Used for categorical data

Bars represent ranges of values (bins)

Bars represent categories or groups

The shape of the distribution is important

Individual category values are compared

Bars touch each other

Bars are usually separated

Check for Understanding

Which type of data is most appropriate for a histogram?

A. Eye color categories
B. Types of materials
C. Temperature measurements
D. Favorite programming language

Show Answer

Correct Answer: C

Histograms are used for continuous numerical data, such as measurements like temperature.

Types of Distributions#

Histograms can reveal the shape of a dataset’s distribution, which helps engineers understand how values are spread across the data.

Some distributions are symmetrical, meaning the left and right sides look similar. Others are skewed, meaning the distribution stretches farther in one direction than the other.

Distribution Type

Description

Key Characteristics

Normal (Gaussian)

A symmetrical, bell-shaped distribution where most values cluster near the center.

Mean, median, and mode are approximately equal. Values decrease evenly on both sides of the center.

Right-Skewed (Positive Skew)

A distribution where the tail extends toward larger values.

Most values occur on the left side, with a few unusually large values stretching the distribution to the right.

Left-Skewed (Negative Skew)

A distribution where the tail extends toward smaller values.

Most values occur on the right side, with a few unusually small values stretching the distribution to the left.

As shown in the figure below, skewed distributions stretch more to one side of the histogram.

Three curves comparing positive skew, a normal distribution, and negative skew, each with mean, median, and mode marked.

Fig. 4.2 Comparison of positive skew, normal distribution, and negative skew. In the normal distribution (center) the mean, median, and mode occur at the same location. In the positively skewed distribution (left) the tail extends to the right and the mean is greater than the median and mode. In the negatively skewed distribution (right) the tail extends to the left and the mean is less than the median and mode.#

Check for Understanding

A normal distribution is commonly described as which shape?

A. Bell-shaped curve
B. Flat line
C. U-shaped curve
D. Random spikes

Show Answer

Correct Answer: A

A normal distribution forms a bell-shaped curve, where most values cluster near the center and fewer values appear toward the extremes.

Check for Understanding

Skewness describes the ________ of a distribution.

A. Symmetry
B. Frequency
C. Measurement units
D. Requirements

Show Answer

Correct Answer: A

Skewness describes whether a distribution is symmetrical or stretched more to one side.

Kurtosis#

Three overlapping curves comparing positive kurtosis, a normal distribution, and negative kurtosis.

Fig. 4.3 Comparison of positive kurtosis, normal distribution, and negative kurtosis. The taller, narrower curve represents positive kurtosis, with heavier tails and more extreme values. The wider, flatter curve represents negative kurtosis, with lighter tails and fewer extreme values. The normal distribution sits between them.#

Kurtosis describes how heavy or light the tails of a distribution are.

The tails of a distribution represent the extreme values in a dataset. When the tails are heavier, the dataset is more likely to contain unusually large or small values (outliers).

Type of Kurtosis

Description

High (Positive) Kurtosis

The distribution has heavier tails, meaning extreme values occur more often.

Normal Kurtosis

The distribution has a typical bell shape with a moderate number of extreme values.

Low (Negative) Kurtosis

The distribution has lighter tails, meaning extreme values occur less often.

Kurtosis helps engineers understand whether a dataset may contain more or fewer extreme values than expected.

Creating a Histogram in Excel#

To create a histogram in Excel:

  1. Select the data you want to analyze.

  2. Go to the Insert tab.

  3. In the Charts section, select Insert Statistic Chart.

  4. Choose Histogram.

Excel will generate a histogram with bins that group the data into ranges.

Adjusting Histogram Bins#

Excel automatically chooses a default number of bins, but engineers may need to adjust the bin settings to better represent the data.

To modify the bins:

  1. Click on the horizontal axis of the histogram.

  2. Right-click and select Format Axis.

  3. In the Axis Options panel, adjust one of the following:

  • Bin Width - controls the size of each interval

  • Number of Bins - controls how many intervals appear in the histogram

Choosing an appropriate bin size helps reveal the true shape of the distribution.

Excel also provides options for:

  • Underflow Bin - groups all values below a specified limit into a single bin

  • Overflow Bin - groups all values above a specified limit into a single bin

These options can be useful when you want to separate extreme values or outliers from the main dataset.

Improving Histogram Readability#

After creating a histogram, you may need to adjust the formatting to make it easier to interpret.

Good histogram formatting includes:

  • Adding a clear chart title

  • Labeling the axes

  • Including units of measurement when appropriate

  • Avoiding unnecessary visual effects such as 3-D charts

Clear formatting helps ensure the histogram communicates the data accurately.

Interpreting Histograms#

When analyzing a histogram, engineers often consider several key questions:

  • Where are most of the values located?

  • How wide is the spread of the data?

  • Is the distribution symmetrical or skewed?

  • Are there any unusually large or small values (outliers)?

By examining these characteristics, engineers can better understand patterns in data and evaluate how a system or process behaves.

Outliers#

When analyzing data, some values may appear very different from the rest of the dataset. These values are called outliers.

An outlier is a data point that lies far outside the overall pattern of the data.

Outliers may appear in visualizations such as histograms, scatter plots, or box plots as values that are clearly separated from the majority of observations.

Why Outliers Occur#

Outliers can appear for several different reasons. It is important to understand why an outlier occurs before deciding how to handle it.

Common causes include:

  • Measurement errors

  • Data entry mistakes

  • Instrument malfunction

  • Unusual or rare real-world events

  • Natural variation in the system being studied

Not all outliers are mistakes. Some represent important observations that may reveal new information about the system being analyzed.

Identifying Potential Outliers - Interquartile Range#

One common way to identify potential outliers is by using the Interquartile Range (IQR).

The interquartile range describes the spread of the middle 50% of the data.

The steps for identifying potential outliers are:

  1. Find the first quartile (Q1).

  2. Find the third quartile (Q3).

  3. Calculate the interquartile range: IQR = Q3 - Q1

  4. Calculate the lower and upper bounds:

    • Lower bound: Q1 - 1.5 * IQR

    • Upperbound: Q3 + 1.5 * IQR

Any data points outside these bounds may be considered potential outliers.

Finding Quartiles in Excel#

Quartiles divide a dataset into four equal parts.

  • Q1 (first quartile) represents the 25th percentile of the data.

  • Q3 (third quartile) represents the 75th percentile of the data.

In Excel, quartiles can be calculated using the QUARTILE.INC function.

General structure: =QUARTILE.INC(array,quart)
- array - the range of data values
- quart - the quartile number to calculate

Quartile

Example Formula

Q1

=QUARTILE.INC(A2:A31,1)

Q2 (Median)

=QUARTILE.INC(A2:A31,2)

Q3

=QUARTILE.INC(A2:A31,3)

Once Q1 and Q3 are calculated, the IQR can be computed using: IQR = Q3 - Q1

This value is then used to determine the upper and lower bounds for potential outliers.

Check for Understanding

A dataset has the following quartile values:

Q1 = 24 Q3 = 32

Using the IQR method, calculate:

  1. The interquartile range (IQR)

  2. The lower bound for potential outliers

  3. The upper bound for potential outliers

Show Answer

First calculate the interquartile range:

IQR = Q3 - Q1 IQR = 32 - 24 = 8

Next calculate the bounds:

Lower Bound = Q1 - 1.5 \(\times\) IQR Lower Bound = 24 - 1.5(8) = 24 - 12 = 12

Upper Bound = Q3 + 1.5 \(\times\) IQR Upper Bound = 32 + 1.5(8) = 32 + 12 = 44

Final answers:

  • IQR = 8

  • Lower Bound = 12

  • Upper Bound = 44

Identifying Potential Outliers - Standard Deviation#

Another common method identifies outliers using the mean and standard deviation.

A common guideline is that values outside the following range may be considered potential outliers:

(4.1)#\[\text{Mean}\space\pm\space 3\space*\space \text{Standard Deviation}\]

This method is sometimes used when data follows a normal distribution.

However, when datasets are highly skewed or contain long tails, the interquartile range (IQR) method often provides more reliable results.

Comparing Outlier Detection Methods#

Histogram of bike trip durations with a strong right skew and a long tail of extreme values.

Fig. 4.4 Histogram of bike trip durations showing a right-skewed distribution with several extreme values. Most trips are short, but a few are much longer.#

Different statistical methods can identify very different numbers of outliers, especially when the dataset is skewed.

In the histogram shown, most bike trips are relatively short, but a small number of trips last much longer. These extreme values appear in the long right tail of the distribution.

When two different outlier detection methods are applied:

  • The Interquartile Range (IQR) method may identify many potential outliers.

  • The Standard Deviation method may identify far fewer outliers.

In one example dataset:

  • The IQR method identified 146 potential outliers

  • The Standard Deviation method identified only 3 potential outliers

This occurs because the two methods measure variability differently.

Method

Key Idea

IQR Method

Focuses on the middle 50% of the data, making it more sensitive to extreme values in skewed distributions.

Standard Deviation Method

Uses the mean and standard deviation, which can be influenced by large values.

Because of these differences, engineers often compare multiple approaches and use engineering judgment when interpreting results.

Handling Outliers#

After identifying a potential outlier, the next step is to decide how it should be handled.

Outliers should not be removed automatically. Engineers must first determine whether the value represents an error or a meaningful observation.

Check for Possible Errors#

Some outliers occur because of mistakes during measurement or data entry.

Ask whether the value is reasonable in the context of the problem.

Examples of possible errors include:

  • Typographical mistakes when entering data

  • Unit conversion errors

  • Sensor malfunctions

For example, if a dataset contains a recorded body temperature of 986 \(^\circ\text{F}\), the value is likely the result of a data entry error.

Investigate the Data#

If the value is not clearly an error, examine the data more closely.

Consider questions such as:

  • What makes this data point different from the others?

  • Does the observation represent a rare but valid event?

  • Could environmental or system conditions explain the value?

Outliers sometimes represent important real-world events rather than mistakes.

Compare Results With and Without the Outlier#

Another useful strategy is to analyze the data with and without the outlier.

Consider the following questions:

  • Do the summary statistics change significantly?

  • Does the distribution shape change?

  • Do the conclusions drawn from the data change?

If the results change dramatically, the outlier may strongly influence the analysis and should be carefully considered.

Why This Matters#

Different outlier detection methods may identify very different numbers of outliers, especially when data is skewed.

For example, one method might identify many values as outliers, while another method identifies only a few.

Because of this, engineers should interpret the results carefully and consider the context of the data rather than relying only on a formula.

Check for Understanding

A dataset contains one value that is much larger than all the other observations. After checking the data collection process, the team determines that the value resulted from a sensor malfunction.

What is the most appropriate action?

A. Keep the value because all data should remain in the dataset
B. Ignore the value but leave it in the dataset
C. Remove the value because it resulted from a measurement error
D. Replace the value with the average of the other data points

Show Answer

Correct Answer: C – Remove the value

Outliers caused by measurement or recording errors may be removed after the issue has been verified. Valid observations, however, should typically remain in the dataset.