Improving Teaching and Learning in Intro Stat Classes with - PowerPoint Presentation

Slide1

Improving Teaching and Learning in Intro Stat Classes with Student Response Systems: pre-Implementation Discussion

Dr.

Kehui

Chen, Assistant Professor

Dr. Nancy Pfenning, Senior Lecturer

University of Pittsburgh Dept. of Statistics

dB_SERC

August 2015Slide2

Summary of Abstract

Proposed developing and conducting use of student response system (“clickers”) for

surveys

and short quizzes

throughout the semester to achieve more effective communication between the instructor and

students

specially

designed

data-generating activities

to improve concept learning in statistics

sampling

interactive

case studies

to help students apply statistical methods to real data

analysisSlide3

Summary of Abstract (continued)

Proposal includes plans for

d

ocumentation, adaptation

to other intro-level statistics

classes

a

ssessment

of

learning gains and attitude improvementsSlide4

Background

Very broad cross-section of close to 1,000 Pitt students take intro stats each semester.

Communication

needed for optimal teaching/learning:

Students

benefit from ongoing awareness of their strengths and weaknesses relative to instructor’s expectations as well as their peers

Instructors

should closely monitor knowledge,

understandingSlide5

General Challenge #1 for Intro Stats Instruction

Large class sizes (typically 80+)

limit

opportunities for..

i

n-class discussions so

students

can gauge self- and peer- understanding

(Typical hindrance to learning: Students inclined to give up because “

e

veryone but me knows what’s going on”)

f

requent graded quizzes so

instructor

can monitor all students’ progress

(Typical hindrance to instructor setting appropriate pace: Unaware of students’ failure to comprehend ideas until they perform poorly on an exam)Slide6

Facilitating Self- and Instructor- Awareness of Students’ Comprehension

Immediate anonymous responses with clickers facilitate:

Students’ gauging their own understanding concurrently with that of peers in a way that saves time AND prevents embarrassment

“Reality check” so instructor knows immediately when a skill or concept requires further explanation (or when it’s time to move on to new material)Slide7

General Challenge #2 for Intro Stats Instruction

Key processes in learning Statistics:

d

esign of experiments/observational studies

data collection

data cleaning

data analysis

Because of time constraints, traditional lecture mainly just focuses on data analysis.Slide8

Clickers to Convey all Key Processes of Intro Stats Instruction

Activities can highlight:

study design:

how to divide class into treatment/control groups, etc.

data collection:

anonymity, sampling bias, which response options to provide, etc.

data cleaning:

possible modification of response files

data analysis:

opportunity for immediate displays and summaries of meaningful data from students themselvesSlide9

General Challenge #3 for Intro Stats Instruction

Common “missing link” in understanding:

Probability and Sampling Distributions

Whereas recitations may be conducted in computer lab, with software available to explore behavior of repeated random samples, lecturers ordinarily can’t gather data on the spot from all class members electronicallySlide10

Clickers to Convey Key Ideas of Probability and Sampling Distributions

Instead of students having to take it on faith that instructor has conjured up multiple random samples, each student participates actively in the process and witnesses patterns as they unfold in real time.Slide11

Proposed Transformations

1.

Surveys

and

Short Quizzes

First-day clicker-based survey of students’ backgrounds

First-day administration of pre-test featuring well-established valid, reliable conceptual questions taken from CAOS* instrument

Subsequent survey about attitudes and ability to self-assess progress

Periodic insertion of questions for students to discuss, then answer via clickers, addressing most challenging concepts as identified by the instructor and experienced TA

Last-day administration of post-test

(* Comprehensive Assessment of Outcomes in a first Statistics course, developed by

delMas

et al)Slide12

Proposed Transformations

2

. Concept Learning via Clicker

Activitie

s (examples)

Sampling Distributions: Generate repeated collection of sample means or

proportions,

explore as a group how patterns evolve, observe differences resulting from modifying population parameter, sample size, etc.

Inference: Generate repeated collection of confidence intervals or hypothesis test

P

-values to best understand how these inference results behave in the long runSlide13

Proposed Transformations

3.

Interactive Case Study

and Peer Instruction

Group data analysis (

eg

. simple regression) by class using

a real data

set:

Pose

multiple-choice questions at each key

step, have students discuss with classmates, then choose

most appropriate way to proceed.

This gives them experience in making appropriate

problem solving decisions on their

own,

and

developing

critical thinking. Slide14

Goals and Assessments

Goal #1: More Effective Communication

Assessment #1:

Survey

questions

will ask students extent to which they agree with statements like

,

“For most of the lectures, I had a pretty good idea of how much my classmates and I were understanding new material.”

Results will be compared with those of students in a non-clicker-based lecturer’s class.Slide15

Goals and Assessments

Goal #2: Increased Engagement

Assessment #2:

Survey questions will ask students extent to which they agree with statements like,

“Participation with clickers improved my understanding of the subject content

,

”

and

“Use of clickers helped me focus and pay more attention during lectures.”Slide16

Goals and Assessments

Goal #3: Improved Conceptual Learning

Assessments #3:

P

re- to post-test:

gains

will be

compared for

items based on “transformed”

material versus items

based on

non-transformed

material.

Midterm and final exams:

performance

will be compared for items

testing

on “transformed” material versus items

testing

on

non-transformed material

. Slide17

Challenges and Responding Plans

To offset additional time taken up by use of clickers, students’ feedback will guide instructor to reduce time on topics that they grasp quickly.

Experienced Graduate

TAs

will be part of the course transformation. The instructor will have frequent communication with recitation TAs to use lecture time more effectively. Slide18

Sustainability, Scalability

Clickers provided by this dB-SERC grant will be available for other Fall 2015 and future classes

Testing of clicker function can be done periodically by students in lecture

Maintenance costs (

eg

. replacement batteries) will be requested from Stats Dept.

Materials developed (

eg

. questions, activities) will be made available to other intro-stats instructorsSlide19

Budget

Clickers:

$

4

0*100

=$4,000

Half-time TA: $7,500

(Note: grad student deemed to be best-suited for the job isn’t available during lecture times. Fortunately, some connectivity can be achieved by having him hired as TA for 2 of the 4 recitations)Slide20

Example 1: Clicker Activity for Understanding Standard Deviation

Display

distribution and discuss standard deviations when students use clickers to haphazardly choose...

between options 1 and

5;

then between options 1, 2, 3, 4,

5;

then between options 2, 3, 4.

(The first will have largest

sd

, the last will have smallest.)Slide21

Example 1 Concept: Understanding Standard Deviation

CAOS Questions #14 & 15:

Five histograms are presented below. Each histogram displays test scores on a scale of

0 to

10 for one of five different statistics classes

.Slide22
Slide23

Example 1 Concept: Understanding Standard Deviation

#14. Which of the classes would you expect to have the lowest standard deviation, and why?

Class A, because it has the most values close to the mean

Class B, because it has the smallest number of distinct scores.

Class C, because there is no change in scores.

Class A and Class D, because they both have the smallest range.

Class E, because it looks the most normal.Slide24

Example 1: Understanding Standard Deviation

#15. Which of the classes would you expect to have the highest standard deviation, and why?

Class A, because it has the largest difference between the heights of the bars.

Class B, because more of its scores are far from the mean.

Class C, because it has the largest number of different scores.

Class D, because the distribution is very bumpy and irregular.

Class E, because it has a large range and looks normal.Slide25

Example 1: Concept for comparison

Comparable CAOS questions on topic not taught with clickers:

#17.

Imagine you have a barrel that contains thousands of candies with several different colors. We know that the manufacturer produces 35% yellow candies. Five students each take a random sample of 20 candies, one at a time, and record the percentage of yellow candies in their sample. Which sequence below is the most plausible for the percent of yellow candies obtained in these five samples?

30%, 35%, 15%, 40%, 50%.

35%, 35%, 35%, 35%, 35%

5%, 60%, 10%, 50%, 95%

Any of the above.Slide26

Example 1: concept for comparison

2nd CAOS Question on Topic not Taught with Clickers:

#22. Researchers surveyed 1,000 randomly selected adults in the U.S. A statistically significant, strong positive correlation was found between income level and the number of containers of recycling they typically collect in a week. Please select the best interpretation of this result.

We can not conclude whether earning more money causes more recycling among U.S. adults because this type of design does not allow us to infer causation.

This sample is too small to draw any conclusions about the relationship between income level and amount of recycling for adults in the U.S.

This result indicates that earning more money influences people to recycle more than people who earn less money.Slide27

Example 2: Understanding Confidence Intervals in the Long Run

CAOS Questions #28, 30, 31

CAOS Questions on topic not taught with clickers: #25, 26, 27Slide28

Example 3: Understanding Hypothesis Tests and P-values in the Long Run

CAOS Question #40

CAOS Question on topic not taught with clickers: #36Slide29

Conclusion

Students and instructors alike will benefit from the implementation of this project made possible by dB-SERC. We look forward to reporting progress during the coming fall semester.

Questions?

Tips on clicker use?

Please c

ontact

nancyp@pitt.edu

khchen@pitt.edu