Suggestions for best practices for statistics education and using statistical software as an educational tool in the classroom v2 10242011

Suggestions for Best Practices for Statistics Education and Using Statistical
Software as an Educational Tool in the Classroom

Simon KING

Abstract

The high school course that should be affected the most from the advent of technology in the classroom
is probably statistics. However, statistics classes mostly use graphing calculators with perhaps the
occasional use of statistical software to give students experience of statistical output. While this
circumstance is often due to school resources and their availability, many schools with computer
resources are not often utilizing statistical software in their statistics classes.

The use of technology and statistics in and away from the classroom is accelerating at a fast pace, and it
would be prudent to keep that pace in the classroom.

This paper recommends best practices for statistics education in the classroom and demonstrates how
to use statistical software as an educational tool in the classroom, including ideas for assessment.

Background

Like many other high school mathematics teachers, my career in teaching statistics happened by
accident; a hole in class scheduling that needed plugging and no-one else in the mathematics
department desiring to teach it. What followed in my first year of teaching statistics was a steep learning
curve where I was working hard to stay ahead of my students and my closest friends were the statistics
textbook and my graphing calculator. Once you survive the first year, you reflect and start planning for a
new year, asking the question, “Now what?”

What becomes obvious in the first year of teaching statistics is the potential of a statistics course; it
allows a teacher to be creative in designing lessons, limited to the time to plan and the resources the
school has available (and often the limited access to them). Probably more than any other high school
course, the teaching and learning of statistics in the classroom is highly influenced by the use of
technology.

Section 1 will recommend best practices in the classroom. In section 2, this paper will specifically outline
how to best use statistical software in the classroom.

SECTION 1 – Suggestions for best practices in the classroom

1.1 Most of the students taught in a statistics class will not be statisticians

At a college level, “Today’s introductory statistics course is actually a family of courses taught across
many disciplines and departments. The students enrolled in these courses have different backgrounds
(e.g. mathematics and, psychology) and goals (e.g., some hope to do their own statistics analyses in

research projects, some are fulfilling a general quantitative reasoning requirement)” (Aliaga, et al., 2010,
p. 10). A high school statistics course gives students opportunities to explore interests through data.
While a statistics teacher need not react and adapt to every social trend, “Data analysis may be a way to
help bridge the gap between what students learn in school and their everyday experiences and concerns
and thus enrich their school experiences.” (Barnes, 2009, p. 614)

1.2 Give students the experience of research and reading of journal papers and articles

It is important that students in a statistics class understand that statistics is an applied discipline. When
having students read papers and articles, it is important to initially provide ones that are well written
and accessible. A paper that is difficult to read deflects the student away from the learning experience
they should be getting. This experience also helps students understand what a scientific paper looks like
before they attempt their own.

It is also important for students to read articles that do not have statistical significance to demonstrate
that these papers still contribute to our understanding of a given topic. Many statistics students will
conduct experiments that will indicate no statistical inference, but on reflection, they should understand
that this is not a ‘failure’ and can often provoke more discussion (e.g. type II error, experiment design
issues, hypothesis validity, etc.).

When the students read papers, they can also learn about ways to test a particular hypothesis. They can
adapt or replicate a test they have researched (see section 1.4). This supports the students in the
development of “statistical thinking”.

1.3 The teacher as a statistical consultant

The GAISE: College Report says that students should know, “When to call for help from a
statistician.” (Aliaga, et al., 2010, p. 13),thus a good finishing point at the end of a high school
statistics course is to have students research, design and conduct an experiment with statistical
inference and write a paper of their findings. The teacher plays the role of the statistical consultant
to advise and guide the process. The following items are crucial in the process:
Before designing and conducting the experiment, the students must have researched journal
articles (or similar) on the topic.
The experimental design needs to be approved by the ‘consultant’.
The ‘consultant’ will support the students in appropriate use of statistical inference.
The scientific method is a ‘continuous loop’ (figure 1). “We ultimately need to engage students
in all phases of the investigative cycle of statistics, including data gathering, data analysis, and
inference” (Groth & N.Powell, 2004, p. 106)

1.4 Use psychology context to teach statistics

Psychology is one example, but in actuality many disciplines (or a blend) could be used to teach statistics
(e.g. business, biology, engineering, etc.). What makes psychology effective is the accessibility of the
subject matter to high school students. Simple exampleswhich students can study and test include
researching and testing the effect of the ‘Mozart Effect’ or chewing gum on concentration, or brand
preference (i.e. “The Pepsi Challenge”).

What the teacher has an opportunity to do is to present the topic of research first and then find the
statistical inferential solution. Wiberg (2009) wrote, “Instead of teaching statistical methods and then
applying them to psychological problems and research questions, the order was swapped. The students
were presented with a psychology research article containing statistics which they would encounter
during the course” (section 4.2)

A good example of this is to have students study the ‘Matching Hypothesis’ which states, “The matching
hypothesis proposes that we don’t seek the most physically attractive person but that we are attracted
to individuals who match us in terms of physical attraction.” (Eysenck, 2009, p. 4). The students read
three journal articles on the topic, which demonstrate the development of the understanding of the
hypothesis over time and how to test for it. These articles help the students understand good
experimental designs to test for the matching hypothesis and just as importantly how some designs
have failed. Usually, the student experimental design solution involves a ranking test which is the
intended statistical learning component.

This promotes the understanding of “statistical thinking”, or students’ understanding of the why and the
how of conducting of statistical experiments. Additionally, they gain experience of the experimental
process.

Figure 1 – The ‘continuous loop of experimental design’

1.5 Promote ‘Statistical Literacy’

A statistics teacher needs to consider what information they want their students to retain in the future -
weeks, months and even years after they have completed the statistics course. Most content, if not used
or reviewed on a regular basis is not retained after completing a course (although reviewing of notes
should recall them), but it is reasonable to suppose that the skill of statistical thinking (ability to think
logically and analytically) stays with the student for much longer. The development of a statistics course
curriculum should have ‘statistical literacy’ as one overarching aim that can be supported through the
course objectives (i.e. content). One technique to promote ‘statistical literacy’ is to have students look at
how the media uses statistics. What is key to this process is having the students appropriately reflect on
what is presented. For example the following prompts might be used to explore the graphic (Figure 2)
(CNN)

What conclusions can you come to about the topic by exploring the media?
How effective is the visual in conveying the important conclusions about the topic?
Has the data been used appropriately? Explain.

Figure 2 – Telephone poll results concerning Terri Schiavo case - CNN/USA TODAY/GALLUP POLL(CNN)

Students also need to be encouraged to explore thehow the data is used objectively versus subjectively.
When students design and conduct their own experiments, they need to approach their analyses and
discussion objectively. They also need to be aware that, in the end “data beat anecdotes”(Aliaga, et al.,
2010, p. 11).

1.6 The statistics teacher ‘living and breathing’ statistics

Outside of the classroom, it is crucial that a high school statistics teacher furthers his interest in the
subject. There are many ways of doing this:

Read “The Lady Tasting Tea: How Statistics Revolutionized Science in the Twentieth
Century” (Salsburg, 2002)
Become a member of the American Statistical Association
Take further courses in statistics (e.g. enroll part-time in M.S. Statistics program)

Statistics teachers enjoy working with students who have curiosity about the world of statistics and love
learning; the teacher can help the students develop these characteristics by role modeling the behavior
they want to see in them.

1.7 Statistics should not be taught like a mathematics course

“In mathematics, context obscures structure. In data analysis, context provides meaning” (Cobb &
Moore, 2000). Many high school statistics teachers come into statistics teaching with a traditional
mathematics background and it can take time to understand that context is fundamental in statistics
education. The teacher also needs to be aware that many students taking a statistics course for the first
time struggle with the transition from a traditional mathematics course to a statistics course. They often
get frustrated with the expectations that they will be writing and that solutions to problems are not just

simply ‘right or wrong’ like in a traditional mathematics course. Students often also struggle initially to
understand the difference between descriptive statistics and inferential statistics.

It might be prudent for a statistics teacher to spend time at the start of the year giving tasks that directly
address these issues. One example of how to do this would be through the exploration of the US
election polltracker (figure 3) (BBC, 2008)

Figure 3 – US Election Polltracker(BBC, 2008)

The 2008 US general election provides context. As an interactive visual, the students can explore the
context of changes in the time graph. The concept of statistical inference can also be explored as margin
of error and sample size is indicated (the students can also explore the idea of a statistical ‘dead heat’).

1.8 Teaching without the textbook

US education in general is tied to the heavy use of textbooks (and the other resources that now come
with them) for learning. Overusing a textbook can often lead a teacher away from using creativity to
design lessons and units. Moving away from teaching with a textbook challenges a teacher to
individualize their approach and pedagogy and ultimately gives them full control of the student learning.
Teaching without a textbook is not recommended for teachers new to statistics education as it takes
time master course content and collect resources.

1.9 Student Feedback

Part of a teacher’s responsibility is to demonstrate to their students how to be a reflective learner. The
most effective way to do this is by obtaining student feedback periodically and making reasonable
changes to the course on sections completed (for future classes) and sections not yet covered. Student
buy-in to this feedback is essential. There are a few steps that make this an effective process:
The student feedback should be anonymous
The student feedback responses are discussed in class
Some changes to the course, however small should be agreed upon with the students as a result
of the feedback
Students must understand they need to be constructive and respectful when giving feedback
The statistics teacher is not defensive about constructive criticism

1.10 Incorporating a multicultural curriculum

A statistics course presents a teacher with a great opportunity to shape the whole student. By using
appropriate datasets, a statistics course can be enhanced as a multicultural curriculum. For example,
exploring a sample of thirty countries from the world prison population data (data: Walmsley, 2007)
gives a right-skewed distribution (figure 4)

Figure 4 – JMP® output of a sample of 30 world prison population data(Data: Walmsley, 2007)

The outlier in this sample is USA with 714 people in prison per 100,000 of national population. To
promote discussion about multi-cultural issues, the students can be asked, “What are the reasons that
USA has 714 people in prison per 100,000 of the national population? Research is encouraged.”
Students need to be encouraged to be respectful and justified in their responses, but the demographics
of the 714 is usually discussed (i.e. race, gender, age, etc.), as well as societal and cultural norms.

1.11 Collecting Datasets

The applications of statistical concepts can often very much depend on the quality of the dataset used.
Whenever possible, real data should be used to give a better context. When a teacher plans a
curriculum there are always difficulties when a teacher needs a certain type of dataset to help students

explore a concept. The best way to approach this challenge is for the teacher to spend a little time each
week looking for, finding and storing datasets. The teacher should also be familiar with the
characteristics of datasets they have collected. So, rather than have to find a new dataset to help with
the application of a new concept, they will instead use one from their dataset library. Datasets can also
be ‘recycled’. That is, datasets can re-used in a course to explore different concepts and their
application. This has the added advantage of the student becoming familiar with the dataset and its
characteristics. The ‘GAISE: College Report’ recommends that a statistics teacher, “search for good, raw
data to use from web data repositories, textbooks, software packages, and surveys” (Aliaga, et al., 2010,
p. 16).

1.12 Statistical Applets

Web-based applets serve a purpose of helping students explore concepts kinesthetically and visually.
However, the statistics teacher needs to be careful in their use. “What these tools often gain in
visualization and interactivity, they may lose in portability. And while they can be freely and easily found
on the Web, they are not often accompanied by detailed documentation and activities for student use.
The time required for the instructor to learn a particular applet/application, determine how to best
focus on the statistical concepts desired, and develop detailed instructions and feedback for the
students may not be as worthwhile as initially believed.” (Chance, Ben-Zvi, Garfiel, & Medina, 2007, p.
7). The ‘Regression by Eye simulation’ (Rice Virtual Lab in Statistics) (Figure 5) is accompanied with
instructions and a short exercise. The simulation is very good both visually and interactively, but the
challenge for the teacher is to link this into learning. Another issue is that it does not use ‘real’ data and
so can be abstract for many students.

Figure 5 – ‘Regression by Eye Simulation’ Applet (Rice Virtual Lab in Statistics)

The applet ‘Understanding the Least-Squares Regression Line with a Visual Model: Measuring Error in a
Linear Model’ (NCTM)(Figure 6) is a wonderful interactive visual to support the student understanding

of least-squares regression. It is accompanied by some very good reflective questions. The only issue is
the lack of real data so some students may struggle with its abstract nature.

Figure 6 – Applet ‘Understanding the Least-Squares Regression Line with a Visual Model: Measuring
Error in a Linear Model’(NCTM)

The applet ‘Simulating the probability of a head with a fair coin’(Webster West) has some instructions
for use. What makes this a very useful educational tool is that it uses a real application as a simulation
(i.e. coin flips) and in that regard a teacher could easily add a few prompts to support the student
discovery of The Law of Large Numbers.

Figure 7 – Applet ‘Simulating the probability of a head with a fair coin’(Webster West)

1.13 Play games to collect student data in-class

Using flash applets and games there are many ways to quickly and efficiently collect and analyze data in-
class. For example, one experiment to conduct in class would be to test for the improvement of
concentration and performance on a task by chewing gum. To collect data for this experiment, all
control and treatment groups would play a concentration game (MathIsFun.com, 2011) and an analysis
conducted (bias and confounding would be part of the discussion). Provided it is a timed task, the
teacher has many appropriate flash games to choose from. What becomes evident on analysis, is how
well the students know game data; in particular variability and the effect of outlying values. The phrase
‘know your data’ comes to mind. However much students enjoy collecting data in this way, the learning
point cannot be lost, however. “In order for fun to avoid being (or perceived as) frivolous or unrelated to
course objectives, it is important to present it with structure and intention.” (Lesser & Pearl, 2008).
There is nothing simpler than generating data through playing an online game in class for use of analysis
(for example, males versus females in a comparison of two means). The GAISE: College Report
recommends “Us*ing+ class-generated data to formulate statistical questions” (Aliaga, et al., 2010, p.
16). While using ‘real data’ is good, having students generate and analyze their own data is one step
better.

Figure 8 – Concentration Memory Game

SECTION 2 – Using analytical software as an educational learning tool

A high school statistics course is usually centered round delivering the College Board Advanced
Placement Statistics curriculum (College Board) and the use of a graphing calculator. According to the
College Board, “Students are expected to bring a graphing calculator with statistical capabilities to the
exam, and to be familiar with this use.” (College Board, 2005). This creates a dilemma for the statistics
teacher who is fortunate enough to have access to technology, specifically statistical software. While the
statistical software presents a wonderful opportunity to enhance the learning experience of the
students, using the software more extensively at the expense of the time students might spend with a
graphing calculator would potentially put the students at a disadvantage in the College Board AP
Statistics examination. Therefore, when statistical software (e.g. JMP®, Minitab®, R) in the classroom is
used, it is mostly to give the students the experience of seeing and interpreting statistical output. This
approach is reflected in most high school statistics textbooks, which heavily incorporate graphing
calculators and usually focus on the use of analytical software at the end of the chapter or as
supplemental material. This is a reasonable approach to take considering the College Board policy on
graphing calculator use in the AP Statistics examination and the challenge of limited resources in many
public schools in USA.The GAISE College Report (Aliaga, et al., 2010, p. 20) indicates, “Regardless of the
tools used, it is important to view the use of technology not just as a way to compute numbers but as a
way to explore conceptual ideas and enhance student learning as well.”

This considered, this section will present examples and ideas for how statistical analytical software can
be used as an educational learning tool beyond solely an analytics tool. This section will specifically give
examples using JMP®, however these examples and ideas can be adapted for other analytical software.

2.1 Supporting the use of statistical software

Careful consideration needs to be taken when deciding what analytical software to adopt for the
classroom. Having students use software that requires programming can deflect student attention away
from the aims of the course. The use of graphical calculators and statistical software moved the teaching
of statistics towards analysis and interpretation and away from pen and paper number crunching. It
could be argued that excessive time spent on programming again distracts from the real purpose of the
course.

It is important that if we are assessing student analytical and reasoning skills and conceptual
understanding, they should not be penalized because they cannot recall how to program a statistical
test or what buttons to press.

Using new software can be challenging. Most software we have learned to use (e.g. web browser, word
processor, etc.) is through experimentation and the occasional use of the ‘help option’. For students
using statistical software we need to follow the same path with perhaps one little addition. For students
to learn new software the teacher needs to consider the following:
Give students time and opportunity to play and explore particularly with graphing
Provide step by step instructions for a given statistical process
Provide short video tutorials (example – figure 8). These should be no more than five minutes
long and tabbed if possible

Figure 8 – example of video tutorial for use of JMP®

2.2 Visuals

“Technology has also expanded the range of graphical and visualization techniques to provide powerful
new ways to assist students in exploring and analyzing data and thinking about statistical ideas, allowing
them to focus on interpretation of results and understanding concepts rather than on computational
mechanics.” (Chance, Ben-Zvi, Garfield, & Medina, 2007, p. 4). Just as technology has changed how we
teach statistics, visuals have evolved such that we can explore data far more effectively. An effective
classroom experience is spending time learning about the history and evolution of visuals (e.g. Florence
Nightingale(Lienhard, 1998 - 2002)), the representation of multiple random variables in one visual (e.g.
bubble plots, Charles Joseph Minard(Wikipedia, 2011)) and the more recent phenomena of interactive
visuals. In addition, having the students focus on what constitutes a good (or bad) visual is time well
spent. For example, Figure 9 below is discussed by studentsand they have to come up with a better
visual recommendation.

Figure 9 – example of poor visual - JMP® output of student test scores

JMP® is a useful product in that it makes one pay particular attention to the classification of the random
variables (Figure 10 - JMP® SE 8.01 – help (SAS®)). If a random variable is wrongly classified, the output is
usually different from what was expected.

Figure 10 - JMP® SE 8.01 – help (SAS®)

Kinesthetic learning about is usually about manipulating physical objects. For many people it is an
engaging way to learn. “I hear and I forget, I see and remember, I do and I understand” (Confucius, 551
B.C.). Pressing buttons on a graphing calculator or computer is not kinesthetic learning

A good example of kinesthetic learning is exploring bivariate data in JMP®. Teachers and students have
an opportunity to exclude points and re-plot the regression line in the same graph. This is an excellent
way to explore influential points (figure 10).

Figure 10 – JMP® Fit Y by X Analysis. Excluding points and re-fitting the regression line.

2.3 Explore Test Assumptions

Statistical software provides an opportunity to explore ideas in more depth. For example, when
discussing the Central Limit Theorem (CLT), a student is usually informed that the sample size should be
at least thirty if the population is non-normal. This, as most statisticians know is a simplification and in
reality one needs to know as much as possible about the distribution of the population. Using JMP® and
a provided script, students can explore different ‘population’ distributions and make discoveries about
what sample size is appropriate. This can lead to a better understanding of what conditions are needed
for the CLT. For example, it can be shown that a sample size of two from a normal distribution (UCLA
Statistics, 2008)provides confidence intervals that capture the mean 98% of the time (Figure 11).

Figure 11 – JMP® script - confidence intervals (sample size of two) from a normal ‘population’

The students also explore non-normal populations to see if a sample of thirty holds. For the learning
process, students reflect on why they got the results they did (percentage of confidence intervals
containing the population mean) and using real data provides better context for them to be able to
understand their results.

2.4 Create and explore visuals not in a traditional high school statistics class curriculum

Using statistical software, students have the opportunity to explore visuals not in a traditional high
school curriculum. Many teachers have students look at and explore interesting visuals online, but rarely
do students create anything other than traditional visuals. Again, analytical software presents new
opportunities for students to create visuals from bubble plots (figure 12) to Pareto plots (figure 13).
Using a Pareto plot, students can explore the ‘80-20 rule’.

Figure 12 – bubble plot of year, median house price and median house income(U.S. Census Bureau, 2009)

Figure 13 – Pareto Plot for auto sales, July 2009 (Motorsales, 2009)

2.5 Teaching content beyond the high school statistics curriculum

Resources such as analytical software open the availability of statistical inference beyond the scope of a
regular high school statistics course. Examples of content that might be added include bootstrapping,
Fisher’s exact test and ANOVA. Perhaps the best rule of thumb regarding what content to add is that the
concepts of any new content added needs to be accessible to student. For example, Fisher’s exact test
needs a good grasp of probability. If the course has a high degree on probability content, then perhaps
this is an option. Another potential option is a more in-depth exploration of normality. Adding a Shaprio-
Wilk goodness-of-fit test add an extra dimension of analysis for students (Figure 14)

Figure 14 – exploring normality

2.6 Exploring large datasets with multiple variables

While multivariate analysis is probably beyond the scope of a first year statistics course, statistical
software provides opportunities for students to explore large datasets with multiple variables. This
mirrors the work of a statistician, who often has to explore covariates in bivariate analysis. Visualization
of multiple variables often paints a very different picture than if we limited ourselves to one or two
variables. For example, when we create a scatter plot of average SAT score versus percentage of
students taking by State (Figure 15) we see what looks like a negative association.

Figure 15 – Bivariate Plot of TOTAL SAT score versus Percent Taking by STATE (College Board, 2001)

However, adding the category of region (in the US) to the scatterplot (Figure 16) paints a different
picture:

Figure 16 – Bivariate Plot of TOTAL SAT score versus Percent Taking by STATE with added indicator of US
region (College Board, 2001)

With a little more investigation by students it becomes clear that there are two populations in one
scatterplot; one population (top left oval) is of the states where college bound students take only the
SAT. The second population (bottom right oval) arethe regions where the students take the ACT as the
primary test and elect to take the SAT.

A correlation matrix (figure 17) explores multiple relationships between different body measurements
and mass. The matrix provides a useful exploration of which body measurements most correlate to mass
in addition to body measurements that correlate to each other.

Figure 17 – correlation matrix of body measurements (data: (SAS ))

2.7 Natural Variability

Variability and its measurement are fundamental to a high school statistics course. Students need to
spend time discovering and identifying natural variability. Key to statistical inference is the
measurement of variability. “Variability is natural, predictable and quantifiable” (Aliaga, et al., 2010, p.
11). Technology can be used to explore this concept, with particular attention paid to visualization. An
example of how students can explore the concept of natural variability is by giving them simulated data
of rolls of dice and coins and they have to determine which of the dice and coins are biased (Figure 18 –
‘Dodgy Dice’ exercise).

Figure 18 – Exercise – ‘Dodgy Dice’

When the students attempt this exercise at the start of the year they explorethe distributions of the
events (Figure 19). What they usually do is compare these distributions to what they think the expected
distributions of the events would look like. The further away the simulated distribution is from their idea
of the expected distribution, the more likely it is that they will say the coin or dice are dodgy. What they
have done is draw ‘a line in the sand’ - one side is natural variation and the other side a biased coin
and/or dice. The students frequently disagree where to draw the line which is a helpful conversation for
when the arbitrary nature of alpha=0.05 is introduced. The second time they attempt this exercise is
applying chi-square goodness-of-fit test when they can accurately measure variability.

Figure 19 – typical student JMP® output from ‘dodgy dice’ exercise

2.8 Explore Distributions

The binomial distribution is mostly taught at high school level as a probability application because of the
type of questions expected in the multiple-choice section of the College Board AP Statistics examination
(College Board). However, using statistical software it can be explored as one of the family of
distributions. Crucially, to have students explore the binomial distribution visually rather than calculate
individual probabilities opens opportunities for students to explore statistical concepts. The Zener Cards
exercise (Figure 20) demonstrates this idea.

Figure 20 – Exercise Zener Cards

The students will obtain output as shown in Figure 21. Part c. of the exercise creates a good discussion;
because of people’s skepticism of psychic abilities, they consider their friend not to have psychic abilities
unless there is there is good evidence otherwise (i.e., null and alternate hypotheses). They need to
decide what the probability outcome needs to be less than to consider their friend to be psychic (i.e.,
alpha value). This is an excellent discussion to review before introducing hypothesis testing.

Figure 21 – typical student JMP® output for Zener Cards exercise

2.10 “Association is not causation” (Aliaga, et al., 2010, p. 11)

Statistical software is an excellent tool for exploring association versus causation. An excellent class
exercise can be taken from the paper and dataset, “Storks deliver babies” (Matthews, 2000). A
multivariate data set here is crucial, to not only explore the association between “pairs of storks” and
“birth rate per year (1000’s/yr)” but to visually explore covariates of these variables to a third variable or
fourth variable. The bubble plot (Figure 22) shows the covariate “Size by Area km²”, while the 3D scatter
plot (figure 23) shows the three previously mentioned variables with the addition of the covariate
“Humans (millions)”. While 2D representations of 3D scatter plots are limiting as visuals, JMP® allows
users to interact with the 3D scatterplot by rotating its axes, etc.

Figure 22 – Bubble plot of “storks deliver babies”

Figure 23 – interactive 3D scatter plot “Storks Deliver Babies”

2.9 ‘The Balancing Act’ – using statistical software purposefully and to assess students

When you look at all high school and first year college statistics courses, the variety in pedagogy and
available resources are vast. This paper concentrates on the use of statistical software as the central
learning tool in the classroom.The GAISE College report (Aliaga, et al., 2010, p. 20) states, “We caution
against using technology merely for the sake of using technology”. Concern remains that using statistical
software becomes ‘too easy’ and students just end up pressing buttons and ‘chasing a p-value’ without
fully understanding where this value comes from. “Menu driven is commonly easier for most users as it
allows the user to navigate using the mouse and to hunt and peck a bit more, which has both
advantages (students don’t feel lost) and disadvantages (often trial and error strategy rather than real
thought when choosing a command).” (Chance, Ben-Zvi, Garfield, & Medina, 2007, p. 5). This highlights
the potential greatest loss of using statistical software;poorly used resources can result in a loss of
conceptual understanding and the mechanics of statistics.

It can be argued that while there is no real gain from students calculating the standard deviation, there
is great benefit in them seeing the formula and being asked to explore its purpose.Sccording to the

GAISE: College Report, “ helps students understand the role of standard deviation as a
measure of spread and to see the impact of individual values on ”(Aliaga, et al., 2010, p. 18).

While examples of the pedagogical use of statistical software have been presented in this paper,
educators are hesitant to explore the use of statistical software because traditional tests that have a
number of procedural skill questions don’t fit with what is happening in the classroom when using
statistical software. The aims of the class therefore need to be revisited. I would propose developing
new aims for the statistics course and to align assessment with those aims: conceptual understanding,
interpretation, and statistical thinking.

“Conceptual understanding takes precedence over procedural skill.” (Burrill & Elliott, 2000, p. 315) In a
traditional statistics course, all too often procedure blurs concept; some students can use formulae get
correct answers, but cannot tell you why they are doing what they are doing.

Using statistical software, concepts and output can be combined to great effect. “Rather than let the
output be the result, . . . it is important to discuss the output and results with students and require them
to provide explanations and justifications for the conclusions they draw from the output and to be able
to communicate their conclusions effectively.” (Chance, Ben-Zvi, Garfield, & Medina, 2007, p. 16)

In this exercise (Figure 24), the students are asked not only to produce output (see Figure 25), but
answer conceptual questions based on the t-distribution and hypothesis testing.

Figure 24 – Exercise – Weight Loss Programs

The followingconcept and application questions that are then asked:

What types of distributions are the four ‘bell curves’ on the right of the output and what is their
relationship with the t-statistic and ‘degrees of freedom’?
Given that the Null Hypothesis for each test is initially true, what does the p-value tell us (hint:
think natural variability)?
As the mean weight loss increases over the four weight loss programs, how and why does this
affect:
o The t-test statistic?
o The p-value?
o The Null Hypothesis?

Figure 25 – typical student output for ‘Exercise – Weight Loss Programs’

Teachers can still have students calculate in classwork and homework, however, if the purpose is to
reinforce a concept, but it is important to have students explore why they are following a certain
procedure. For example, to help students think about z-scores we can use the following prompts:

What does a z-score measure? Include a sketch to help explain.
For the z-score formula, what is the purpose of the numerator and denominator?
If a z-score of 1 equals a p-value of 0.84 and a z-score of 2 equals a p-value of 0.975, then does a
z-score of 1.5 equal (0.84+0.975)/2? Give your answer and explain your reasoning (a sketch
would be useful)

Classwork, homework and assessments need to be aligned. Tests can be given where the students do
not use a calculator, but rather answer only conceptual questions and statistical output is used
conceptually and interpretively.

Nothing should stop the statistics teacher from showing students statistical tables. Linking a normal
distribution table and the formula for the Gaussian function and calculating z-scores and p-values in
Wolfram Alpha (Wolfram Alpha LLC, 2011) in figure 26 has merit in the understanding ideas, history and
concepts of the normal distribution. However, as the scientific calculator took away the need for
logarithm tables so probably should technology remove the need for normal distribution tables.

Student projects, experiments and papers serve an important purpose (see 1.2, 1.3 & 1.4) of developing
statistical thinking. If at the end of the year, students are expected to research, design experiments,
collect data and write-up their results with the teacher supporting as a statistical consultant, then
gradual learning steps need to be added throughout the year with more support and structure with the
intention of gradually removing the stabilizer wheels. Most students have little previous statistical
experience at the start of a high school statistics course.

SUMMARY

Section 1 gives ideas for best practice for statistics teachers, but it is not an exclusive list. There are
many other suggestions for best practice that have merit, such as supporting ‘active learning’ in the
classroom and giving students frequent feedback on how they are doing.

Section 2 demonstrates that statistical software can be an integrated learning tool beyond just an
analytic tool in the classroom that will help high school statistics education evolve and keep pace with
the practice of statistics in business and research. However, there are many barriers to teachers
adopting such software as a central pedagogical tool. These barriers include the following:

limited access to statistical software in many schools
textbooks having graphing calculators as the central statistical tool
the learning curve required by a teacher and the students in order to use statistical software
‘re-visioning’ of class objectives and assessment

The College Board® expecting all students to sit the Advanced Placement Statistics examination
with a graphing calculator (College Board, 2005).

Integrating statistical software potentially changes the focus of a statistics course and student
assessment, resulting in a shift in conversation in the classroom. Conversations in class are more
frequently centered about statistical concepts and their application, with less time spent on process and
mechanics.

Figure 26 – p-value calculation in WolframAlpha(Wolfram Alpha LLC, 2011)

References

Aliaga, M., Cobb, G., Cuff, C., Garfield, J., Gould, R., Lock, R., et al. (2010). GAISE: Guidelines for
Assessment and Instruction in Statistics Education: College Report. American Statistical
Association.

Barnes, P. A. (2009). Empowering Students through Data. Mathematics Teacher, Vol. 102, No. 8 P614-
620 April.

BBC. (2008, November 8). US Election Polltracker. Retrieved 10 15, 2011, from BBC:
http://news.bbc.co.uk/2/hi/in_depth/629/629/7360265.stm

Burrill, G. F., & Elliott, P. C. (2000). Thinking and Reasoning with Data and Chance. National Council of
teachers of Mathematics.

Chance, B., Ben-Zvi, D., Garfield, J., & Medina, E. (2007). The Role of technology in Imporving Student
Learning of Statistics. Technology Innovations in Statistics Education, 1(1).

CNN. (n.d.). Retrieved 10 15, 2011, from http://mschindler.com/2005/03/22/partisan-shmartisan/

Cobb, G. W., & Moore, D. S. (November 1997). Mathematics, Statistics, and Teaching. The American
Mathematical Monthly, Vol. 104, No. 9, 801-823.

College Board. (2001). www.collegeboard.com. Retrieved 10 15, 2011, from www.collegeboard.com:
www.collegeboard.com

College Board. (2005). Calculators on the AP Statistics Exam. Retrieved 10 15, 2011, from
apcentral.collegeboard.com:
http://apcentral.collegeboard.com/apc/members/exam/exam_information/23032.html

College Board. (n.d.). Statistics Course Description. Retrieved 10 15, 2011, from
apcentral.collegeboard.com: http://apcentral.collegeboard.com/apc/public/repository/ap-
statistics-course-description.pdf

Eysenck, M. W. (2009). A2 level Psychology. Psychology Press.

Groth, R. E., & N.Powell, N. (2004). Using Research Projects to Help Develop High School Students'
Statistical Thinking. Mathematics Teacher, 106-109.

Kaplan, D. (2007). Computing and Introductory Statistics. Technology Innovations in Statistics Education,
1 (1).

Lesser, L. M., & Pearl, D. K. (2008). Functional Fun in Statistics Teaching: Resources , Research and
Recommendations. Journal of Statistics Education, Volume 16, Number 3.

Lienhard, J. H. (1998 - 2002). Nightingale's Graph. Retrieved 11 1, 2011, from Engines of our Ingenuity:
http://www.uh.edu/engines/epi1712.htm

MathIsFun.com. (2011). Concentration Memory Game. Retrieved 10 15, 2011, from
www.mathisfun.com: http://www.mathsisfun.com/games/memory/index.html

Matthews, R. (2000). Storks Deliver Babies. Teaching Statistics, Vo. 22, No. 2, pages 36 - 38.

Motorsales. (2009). www.motorsales.com. Retrieved 6 6, 2009, from www.motorsales.com:
www.motorsales.com

NCTM. (n.d.). Understanding the Least-Squares Regression Line with a Visual Model: Measuring Error in
a Linear Model. Retrieved 10 15, 2011, from www.nctm.org:
http://www.nctm.org/standards/content.aspx?id=26787

Rice Virtual Lab in Statistics. (n.d.). Regression by Eye. Retrieved 10 15, 2011, from Rice Virtual Lab in
Statistics: http://www.ruf.rice.edu/~lane/stat_sim/reg_by_eye/

Salsburg, D. (2002). The Lady Tasting Tea: How Statistics Revolutionized Science in the Twentieth
Century. W. H. Freeman / Owl Book.

SAS . (n.d.). JMP-SE 8.01 - Body Measurements.jmp.

SAS(r). (n.d.). JMP-SE 8.01 - Help.

U.S. Census Bureau. (2009). Income. Retrieved 10 15, 2011, from www.census.gov:
http://www.census.gov/hhes/www/income/income.html

UCLA Statistics. (2008). SOCR Data Dinov 020108 HeightsWeights. Retrieved 10 15, 2011, from SOCR
Statistics: http://wiki.stat.ucla.edu/socr/index.php/SOCR_Data_Dinov_020108_HeightsWeights

Walmsley, R. (2007). World Prision Population List. Retrieved from International Center for Prison
Studies: http://www.prisonstudies.org/info/downloads/world-prison-pop-seventh.pdf

Webster West, R. (n.d.). Simulating the probability of a head with a fair coin. Retrieved 10 15, 2011,
from http://www.stat.tamu.edu: http://www.stat.tamu.edu/~west/ph/probsim3.html

Wiberg, M. (2009). Teaching Statistics in Integration with Psychology. Journal of Statistics Education,
Volume 17, Number 1.

Wikipedia. (2011, 7 9). Charles Joseph Minard. Retrieved 11 1, 2011, from Wikipedia:
http://en.wikipedia.org/wiki/Charles_Joseph_Minard

Wolfram Alpha LLC. (2011). WolframAlpha. Retrieved 10 16, 2011, from WlframAlpha:
http://www.wolframalpha.com/

Suggestions for best practices for statistics education and using statistical software as an educational tool in the classroom v2 10242011

Suggestions for best practices for statistics education and using statistical software as an educational tool in the classroom v2 10242011

More Related Content

What's hot (20)

Similar to Suggestions for best practices for statistics education and using statistical software as an educational tool in the classroom v2 10242011 (20)

Recently uploaded (20)

Suggestions for best practices for statistics education and using statistical software as an educational tool in the classroom v2 10242011