Beruflich Dokumente
Kultur Dokumente
Hollylynne Lee
Friday Institute for Educational Innovation
NC State University
Recall that inferential reasoning involves using information or data available (a sample) to make some claim or
generalization about something larger, or beyond, the data itself. While formal techniques (e.g., t-test, ANOVA,
chi-square test, bootstrapping) can be used for more sophisticated learners, building the habits of inferential
reasoning are a crucial aspect of learning to do statistics, from the very get-go with early learners.
To make inferences from data, it is necessary to be able to say something about how data is distributed. In
statistics, we are continually examining how data is distributed for certain variables, and seeking to understand
and model how to describe the signals in that data that may be a result from some noisy process (Konold &
Pollatsek, 2002). Thus, we want to be able to characterize group tendencies. However, Wild (2006) cautions
us against a curriculum filled with describing univariate distributions and describes how comparing groups is
more interesting as we naturally wonder about how things change over time, in different settings, or with
different groups of people.
“Teaching about the features of distributions for beginners tends to be in the context of a single variable,
that is, in a univariate setting. All too often this has led to students being fed, year after year, a constant
diet of univariate data and contrived univariate situations. I plead with teachers to move on to multivariate
notions such as comparisons between groups and relationships between variables as soon as the most
basic foundations have been laid. This is necessary to avoid infecting students with the dread disease
univariatitis which is notorious for causing its victims to experience sensations of drowning in irrelevance
and, ultimately, death by boredom.” (p. 19)
Indeed, comparing groups has become a common recommendation to assist learners in engaging in question
posing, using more sophisticated data analysis to describe group propensities (e.g., proportion of cases in
certain ranges, measures of center and variability), and providing opportunities for inferential reasoning.
Students have a natural curiosity to wonder about whether one group is better or worse, or just merely different
than, other groups. Thus, this naturally leads to opportunities for examining a distribution for a particular
variable, and then separating that variable to examine differences among subgroups in a sample. This is
sometimes called disaggregated data, where
the same variable is examined across different
subgroups, such as gender, education level, or
a specific categorical variable that makes sense
for the context. For example, noticing the
variability in the distribution of the average
teacher salary in the 50 US States and
Washington DC may make us wonder if we can
expect the same variability in salaries across
different regions of the US. (Data from NEA, 2015)
Page 1
eaching Statistics Through Inferential Reasoning MOOC for Educators,
Lee, H. S. (2017). Using group comparisons to support inferential reasoning. In T
Friday Institute for Educational Innovation: NC State University, Raleigh, NC. Retrieved from:
https://fi-courses.s3.amazonaws.com/tsir/unit%202/TSIRUnit2RollerCoasterInvestigationHandout.pdf
Separating data into regional subgroups and comparing
these distributions can provide an opportunity to make an
inference about what could be expected for salaries if you
are an educator in various regions. The graph indicates
that salaries in the Northeast tend to be much higher than
in other regions, with about 75% of all Northeast states
being higher than 75% or more of the salaries from states
in each of the other regions. However, we can see it is
not true for salaries from all states in the Northeast. Thus,
we would expect higher salaries in general, but cannot be
certain that a particular salary within a district or state would be much better than in other regions in the US.
Research has shown that describing a distribution of data in the aggregate where one characterizes whole
group tendencies, such as center and variability,
is a tough concept for early learners of statistics.
If students are asked questions where they have
to make some claim about how two or more
groups are similar or different, or which group
may be better or worse, then this seems to
encourage students to actually have to describe
distributions in more holistic terms because they
have to make comparative statements. This is
why Level B (see middle row within each phase)
in the SASI framework emphasizes comparison
of groups, attending to variability, and describing
distributions using proportional reasoning. We
need to provide ample opportunities to compare
groups using multiple statistical measures and
graphical representations to help learners
transition from a Level A sophistication to Level C sophistication. When students are ready for more
sophisticated ways of comparing groups, they can engage in more formal techniques such as randomization
tests for the difference in means, or t-tests when the sample assumptions are met.
As seen in the example provided above, comparing groups tasks are not restricted to only comparing two
groups. Students are quite capable, and should be exposed to, situations where they are comparing more than
two groups. These informal comparisons using group propensities, measures of central tendency, measures of
variability, and noting outliers, can help pave the way for more sophisticated approaches used later such as
using analysis of variance (ANOVA).
Using multivariate datasets and having students collect data where they must design data collection, or
experiments, to seek to understand if there are differences between groups (or conditions) are critical
experiences for promoting inferential reasoning through comparison of groups.
Page 2
eaching Statistics Through Inferential Reasoning MOOC for Educators,
Lee, H. S. (2017). Using group comparisons to support inferential reasoning. In T
Friday Institute for Educational Innovation: NC State University, Raleigh, NC. Retrieved from:
https://fi-courses.s3.amazonaws.com/tsir/unit%202/TSIRUnit2RollerCoasterInvestigationHandout.pdf
References
Konold, C., & Pollatsek, A. (2002). Data analysis as the search for signals in noisy processes. J ournal for
Research in Mathematics Education, 33(4), 259-289.
ustralian Mathematics
Makar, K. (2013). Predict! Teaching statistics using informal statistical inference. A
Teacher, 69(4), 34.
Pfannkuch, M., Regan, M., Wild, C., & Horton, N. J. (2010). Telling data stories: Essential dialogues for
comparative reasoning. Journal of Statistics Education, 18(1), 1-38.
tatistics Education Research Journal, 5(2), 10-26. Available
Wild, C. (2006). The concept of distribution. S
http://iase-web.org/documents/SERJ/SERJ5(2)_Wild.pdf.
Wild, C. J., Pfannkuch, M., Regan, M., & Horton, N. J. (2011). Towards more accessible conceptions of
statistical inference. Journal of the Royal Statistical Society: Series A (Statistics in Society), 174(2), 247-295.
Zieffler, A., Garfield, J., delMas, R., & Reading, C. (2008). A framework to support research on informal
inferential reasoning. Statistics Education Research Journal, 7(2), 40-58.
Page 3
eaching Statistics Through Inferential Reasoning MOOC for Educators,
Lee, H. S. (2017). Using group comparisons to support inferential reasoning. In T
Friday Institute for Educational Innovation: NC State University, Raleigh, NC. Retrieved from:
https://fi-courses.s3.amazonaws.com/tsir/unit%202/TSIRUnit2RollerCoasterInvestigationHandout.pdf