Tutor information

If you work in the learning labs or know enough Statistics to be helping a friend in MATH 1342, here is some helpful information.

Overview

We have a new textbook for our statistics course, which will be used in all MATH 1342 classes beginning in Fall 2018. Minitab will still be available to students and staff during at least the first year of this transition, but most or all of the software use will be on a free web-based site called StatKey, created for this textbook (with help and the datasets) by the authors.  Do not start describing how to use Minitab to students unless they specifically ask for that help.  We are encouraging instructors to provide software output for topics that are not covered in Statkey, rather than require students to learn to use additional software.

Chapters 3 and 4 are the biggest deviations from the traditional approach we took with the Moore book. If a student is in one of these chapters, don't assume most of the techniques you picked up in the past (e.g. standard error formulas, test statistics, theoretical distributions) will be relevant here. Instead, spend some time using Statkey to create Bootstrap confidence intervals, and Randomization distributions for hypothesis tests before you can assume you will be able to help students in these chapters.

You will find it worthwhile to explore Statkey for yourself for at least a couple of hours before helping students with it. This software makes it easy to generate comparative graphs and summary statistics, and to explore inference from a re-sampling point of view. We believe the resampling approach helps students form a solid foundation of the ideas of statistics before they use formulas.

The authors of the text provide three 20-minute videos to help teachers / tutors / others who know statistics understand this new approach. We encourage each of you to view these videos. Find them from the “Videos” link at http://www.lock5stat.com/. They are the first three videos listed (called “Camtasia videos”) about Randomization Tests, Bootstrap Intervals, and StatKey.

Resources available to anyone:

Resources available to ACC tutors only:

Increasing emphasis on concepts and the overview

Students taking a first statistics course have usually found it difficult to absorb the basic ideas of statistical inference. Even many of the students who can handle the formulas have only a sketchy understanding what is actually being done in statistical inference and then do not develop facility in thinking about how to obtain information from data.

In statistics education circles, introducing statistics from a re-sampling approach (using simulation and permutation techniques) is considered more effective than the usual theoretical-distribution methods in teaching students to obtain useful information from data. In this way, students learn the main ideas before dealing much with formulas and then continue to focus on those main ideas throughout the course.  

While tutoring a student, on any particular problem/section it is important to start with explaining this particular material fits into the main ideas before starting into any description of calculations.   Over the last few years we have been scaling back how many "by hand" calculations students are asked to do in the formulas and that is still true.  The text does have examples showing how to do "by hand" calculations, but we do not ask students to do all of those.  Before you begin telling a student how to calculate anything, ask the student to show you how their teacher told them they were supposed to calculate it -"by hand" from a formula or by using software. If it is with software, ask whether they used StatKey or something else before starting to explain anything.   

The notes below give suggestions for talking with students about specific places in our text that might be different from what you expect. 

Content-specific notes

Chapter 1

Notice that students are not expected to produce random samples or produce diagrams of experimental designs, as they did in our previous textbook.   

Chapter 2

For descriptive statistics, the emphasis is much more on reading StatKey output than on making any graph or computing statistics by hand. Dotplots are emphasized as a quick way to make a frequency graph of quantitative data. Stemplots (which we had been covering in our previous book) aren’t covered here at all. Comparative graphs are emphasized much more than in our previous materials. Notice the following:

Section 2.1

In section 2.1, see how StatKey deals with categorical variables, particularly the relationship between two categorical variables. 

Section 2.4

In section 2.4, notice how StatKey makes it easy to handle comparative graphs of quantitative variables. (Also notice that it requires the dataset be “stacked.” See the description of such datasets in Sec. 1.1 and notice that all the dataset files for the text are “stacked.”)

Section 2.6

In section 2.6, notice that the material on descriptive regression is less computational and more conceptual than that in the previous text. Calculation of regression coefficients by hand using formulas is omitted, and interpretation of R-squared doesn’t come up until chapter 10, in regression inference.

Chapter 3

Throughout Chapters 3 and 4, students are working with statistics and parameters of various types in the same section. This is a challenge for students. The organization of the StatKey menus helps reinforce which types of variables result in which parameters and statistics. Help students identify what clues in the statement of the problem help them identify the relevant statistic(s) and parameter(s).

Section 3.1

Sampling distributions are difficult to understand for students in all beginning statistics courses. Here, however, there is no need to spend much attention on this section, because all the rest of chapters 3 and 4 are based on using the bootstrap or randomization distribution as an approximation of the relevant sampling distribution. They will learn to “see” the relevant sampling distribution more easily using this method than in the usual methods using theoretical distributions.

We emphasize the idea of an entire population from which we take a sample, and then that the distribution of the statistic has smaller variability as the sample size increases. This latter idea comes up only here, because the bootstrap distribution and the randomization distribution are based on one sample and it has only that one sample size.

Sections 3.3 and 3.4

If students need help with these sections, the tutor will need to understand how to produce bootstrap distributions, by hand and with StatKey, in order to be helpful.   

Chapter 4

Sections 4.1-4.3

These are the basic ideas of hypothesis testing.  Students find the p-values by counting dots in randomization distributions.  (Or by having the software count the dots.)   You will be more comfortable tutoring students with this if you have first done enough of section 4.5 to be able to use StatKey to produce the randomization distributions.   Students are not asked to produce the randomization distributions in these sections, but they are expected to learn here that a randomization distribution is centered at the numerical value in H0.     

Section 4.5

In randomization tests we do not expect the student describe how the randomization is done for each type of test. Most teachers will not expect students to do any of the problems that require students to describe how the randomization is done, although some teachers may expect that for one or two types of tests where the method is fairly simple to explain.

The difficulty is to have the method both “use the data” and be “based on H0 being true.”   There is quite a bit of depth in working out these methods for some of the types of problems we cover, and it is very confusing if students start to get into that depth at this point.  

If you feel very convinced that you need to learn to do at least one of them, start with a test of whether two population means are equal, and the data were collected by random assignment of subjects to two treatments.   

In that case, the appropriate randomization method is to take your entire dataset and randomly re-allocate the entire set of observations to the two treatments.    

That clearly fits "use the data."  It also is based on the H0 claim that the two populations have equal means.

Chapter 6

In the use of the theoretical distributions to do inference on means and proportions teachers are expected to not require students to do all SE (standard error) calculations by hand with a calculator. How many are required will vary among teachers.

Chapters 7-10

In the material on chi-squared tests, ANOVA, and inference on regression some of each is required, but teachers have some amount of choice of which sections to cover and how many details of the calculations to expect the students to learn to do “by hand.”