Slayter on planning quant design for flc projects - may 2011

20
On planning your quantitative research design for a faculty learning community project Elspeth Slayter, Ph.D. School of Social Work Salem State University

Transcript of Slayter on planning quant design for flc projects - may 2011

Page 1: Slayter   on planning quant design for flc projects - may 2011

On planning your quantitative research design for a faculty learning community project

Elspeth Slayter, Ph.D.School of Social WorkSalem State University

Page 2: Slayter   on planning quant design for flc projects - may 2011

Before you begin…don’t read these slides unless you…

• …Have skimmed a basic program evaluation text• …Know what your “intervention” will be

– Example: Increasing student engagement in research courses

• …Have framed your research question– Example: Does the use of a new type of pedagogical framework

increase self-reported student engagement at mid-semester?

• …Are pretty sure you want to compare groups– Example: Pre/post tests on one class, comparison of sections

Slayter 2

Page 3: Slayter   on planning quant design for flc projects - may 2011

When babies are on the way, we prepare a nursery, yes?

• You must prepare a data analysis plan before the data are collected – and eventually – arrive!

• Mistake #1: Create survey, analyze data later, can lead to many problems – don’t do it!

Slayter 3

Page 4: Slayter   on planning quant design for flc projects - may 2011

Metaphor: Preparation of nursery – Preparation of data

analysis plan• A somewhat iterative process

in choosing the following:– Over-arching research question

– Project goals

– Project objectives

– Data collection plan

– Identification of analytic task

– Choice of statistical tests

Slayter 4

Page 5: Slayter   on planning quant design for flc projects - may 2011

Presentation Overview:How to approach this task

5Slayter

Page 6: Slayter   on planning quant design for flc projects - may 2011

Step 1: Identify project research question,

goals, objectives, intervention• Example research question: How can I foster

student engagement in required research courses?– Example goal: To improve student engagement by

changing my teaching framework as explicated to students

– Example objective: To improve class participation grades after infusing new teaching framework in lecture, discussion, assignments, activities

Slayter 6

Page 7: Slayter   on planning quant design for flc projects - may 2011

Step 1: Identify project research question,

goals, objectives, intervention

Slayter 7

Your goal is LINKAGE!!!

Page 8: Slayter   on planning quant design for flc projects - may 2011

Step 2: Design data collection approach that

best answers your question

• Methods of assessing outcome must be best possible/feasible way to answer question– Will numbers or words answer the over-arching

question best?• Maybe quantitative?

– Maybe pre-post-test? Maybe not?

• Maybe qualitative? (if so, skip this presentation!)

Slayter 8

Page 9: Slayter   on planning quant design for flc projects - may 2011

Step 3: Define analytic task, choose

statistical test, example 1

Slayter 9

Page 10: Slayter   on planning quant design for flc projects - may 2011

Step 3: Define analytic task, choose

statistical test, example 2

Slayter 10

Page 11: Slayter   on planning quant design for flc projects - may 2011

Step 3: Define analytic task, choose

statistical test, example 3

Slayter 11

Page 12: Slayter   on planning quant design for flc projects - may 2011

On choosing statistical tests for bivariate analyses

Slayter 12

Page 13: Slayter   on planning quant design for flc projects - may 2011

Requirements for each bivariate statistical test differ

• Comparing groups is one analytic task

• What you are comparing them ON is what matters with this choice

Slayter 13

Page 14: Slayter   on planning quant design for flc projects - may 2011

Link variable type to appropriate statistical test for analytic purpose

Slayter 14

Page 15: Slayter   on planning quant design for flc projects - may 2011

Check assumptions that must be met to conduct statistical testing

• Each statistical test is based on assumptions

• Must make sure assumptions are met or test results are spurious

• Common assumptions: Sample size-related, distribution-related, structure of variable

Slayter 15

Page 16: Slayter   on planning quant design for flc projects - may 2011

Cheat sheet for choosing bivariate statistics

Analytic purpose you are seeking help for:

Test name: Basic requirements: How to interpret:

Comparing 2 or more groups on a percentage/rate

Chi-square test Variable to be compared is nominal

Interpret p value for statistical significance only

Comparing 2 groups only on a percentage/rate

Odds ratio test Variable to be compared is nominal

Interpret odds (1.23 times more likely, 0.40 = 60% less likely)

Comparing mean difference between 2 groups

Independent samples t-test

Variable to be compared is continuous

Interpret p value for statistical significance only

Comparing mean difference between pre and post in one group

Paired samples t-test

Variable to be compared is continuous

Interpret p value for statistical significance only

Comparing mean difference between 2 or more groups

Analysis of Variance (ANOVA)

Variable to be compared is continuous

Interpret p value for statistical significance only, conduct post-hoc test to identify divergent group

Slayter 16

Page 17: Slayter   on planning quant design for flc projects - may 2011

Step 4: Envision finished product

• Make sure your data analysis plan makes sense – is this what I want when I am done?

• Consider using a “table shell”– Tables with no data in them – and a process– Take each question and think about how you will

use it analytically– Fill in which test you will use so you know what

you are doing ahead of time (creating a recipe for yourself)

17Slayter

Page 18: Slayter   on planning quant design for flc projects - may 2011

Step 4: Envision finished product

• Use table shells to clarify thinking pre-data collection

Table 1: Comparison of student engagement rates between class sections

Variable: Target section:N=33

Comparison section:N=31

Statistical test result:

Self-reported engagement score#

Mean =3.5 Mean =2.5 t=2.45* (Independent samples t-

test result)

% reporting increased engagement at end of semester

67% 32% OR=3.42** (Odds ratio test result)

Χ2=3.46*** (Chi-square test result)

# - Likert scale indicates 1=not engaged to 5=totally engaged*** p<.001; **p<.01; *p<.05

Slayter 18

Page 19: Slayter   on planning quant design for flc projects - may 2011

Step 4: Envision finished product

• Checklist once table shells are complete:Do I ask all necessary questions to get my data

into the shape I envision here?Do I ask questions that give me answers in the

correct format?Nominal vs. continuous variable for appropriate

statistical test to meet analytic purpose

Slayter 19

Page 20: Slayter   on planning quant design for flc projects - may 2011

You are good to go!

• Check out my mid-term FLC portfolio from AY 10-11 for table shell examples

• If all else fails, get in touch with questions – [email protected] x7459

• Good luck with your projects and have fun!

Slayter 20