How to conduct a good research study

1
How to conduct a good research study

• STA 309H
• January, 27, 2004

2
Outline

• Group projects
• Group assignments, important dates, etc.
• Project guidelines
• Research methods (what to do and what NOT to do)
• Stop me if you have questions!

3
Group Projects

• See handout for your group assignments
• Presentation dates
• Groups 1 to 4 April 29th
• Groups 5 to 8 May 4th
• Groups 9 to 12 May 6th
• See group project FAQ section on course website
  for other administrative questions

4
My role

• Consult with groups on your projects
• Where to meet
• My office, Seay Psychology Building 3.426H
• When to meet
• Email me to make an appointment (the earlier the
  better)
• EVERY group member should be present when meeting
  with me

5
Steps for conducting research

• Identify a research question

6
Choosing a good research question

• Group project involves conducting an
  investigation of an interesting and original
  question related to one or more themes of this
  class.
• Your project will be evaluated on its
  originality, educational value, and overall
  quality.
• No specific constraints on the subject matter

7
Choosing a good research question

• Interesting
• Wow, I cant wait to tell some of my smartest
  friends about this one!
• Not obvious (Will your findings be surprising and
  exciting?)
• A question where you look at conditions under
  which
• Doable (in your time frame)

8
Choosing a good research question

• Past projects include
• Can experts predict the stock market?
• Is racial profiling used by airport security?
• Should fast-food consumers drive-thru or walk-in?
• What is the density and variability of raisins in
  Raisin Bran cereals?
• Does confidence affect test performance?
• Why do some like Leno while others like
  Letterman?
• What is the optimal strategy in rock, papers,
  scissors?
• What is the probability of transmitting HIV in
  the boxing ring?
• Do blondes have more fun?
• Do winning sports teams improve the citys
  economy?
• Do good liars make good leaders?

9
Research questions to AVOID!!

• In general, try to avoid questions about
• Grades and salaries
• Gender differences
• Comparing students across major and colleges

10
Choosing a good research question

• It is better (but not required) to have a theory
  and a hypothesis about your research question
• Hypothesis
• What do you think is the answer to your research
  question? (You should not KNOW the answer to your
  question, but you may have a hunch or a good
  guess)
• Theory
• Why do you think your hypothesis is true?

11
Steps for conducting research

• Identify a research question
• Conduct a study/investigation to test the
  research question

12
Testing the Research Question

• Analyze data that is already available
• Collect your own data

13
Analyzing data that is already available

• Analytical projects (integration of
  statistical/probability information from
  different sources)
• Transmitting HIV in the boxing ring
• Analyzing data from an existing data source (a
  database)
• Stock market data
• Sports statistics
• Crime data

14
Collecting your own data

• Surveys
• Collect data from the population without
  manipulating or controlling anything,
• E.g. Political attitudes survey
• Experiments
• Collect data from the population by manipulating
  something (an independent variable)
• E.g. drug studies

15
Surveys

• A lot in common with archival/database studies
• Difference you collect the data with a survey,
  while the data is already there with a database
• Similarity both yield observational data (you
  observe the world without intervening). You
  cannot determine any causal relationships.

16
Surveys

• Common things to look out for
• Random sampling?
• Non-response error?
• Wording of questions?
• Social desirability/self-reports?
• Does association causation?
• If there is a relationship between A and B, does
  that mean that A causes B?

17
Surveys

• Coffee may make older lovers perky In a
  survey of people 60 and over
• 62 of coffee drinkers were sexually active,
  while only 37 of non-coffee drinkers were
  sexually active
• Is there a causal relationship between coffee and
  sexual activity?

18
Surveys

• Difficult to know which variables have a causal
  influence on which other variables
• Need an experiment to find a causal relationship
• You manipulate a particular variable and measure
  its influence on another variable or group of
  variables

19
Surveys

• Nevertheless there is a lot of survey and
  database (archival) research out there
• For some research questions, researchers are not
  concerned about causality
• Ethical issues may prevent the collection of
  experimental data (e.g. physical or sexual abuse)
• People have tried controlling for outside factors
  (matched samples design)

20
Experiments

• Researcher manipulates a feature of a task called
  the INDEPENDENT VARIABLE and determines its
  effect on the DEPENDENT VARIABLE
• Independent variable manipulated (this is the
  factor that we think is causing something else)
• Dependent variable measured (we hypothesize
  that the independent variable affects this
  variable- we measure to see if there is an
  effect)

21
Experiments

• At least 2 groups of subjects are treated exactly
  alike in all ways except one (the independent
  variable). Any differences in the observed
  behavior of the two groups are attributed to
  (caused by) the differences in the independent
  variable.
• Experimental group undergoes the manipulation
  hypothesized to have an influence on the
  dependent variable (e.g. the drug of interest)
• Control group receives a different or no
  manipulation this group is used to make certain
  that the result is due to the hypothesized cause
  (e.g. no drug)

22
Experiments

• Random assignment to the groups
• VERY IMPORTANT!!!!
• Why?
• What do you mean by treated alike in all ways
  except one?

23
Experiments

• Drug experiment experimental group gets the
  drug, control group gets nothing is this good
  enough?
• As this study is currently designed, the
  experimental group gets a tablet AND the medicine
  in the tablet (two things are being varied at the
  same time)
• If there is an effect of the drug on health or
  pain (or whatever the DV is), we have a confound
  was the effect due to the tablet or the
  medicine in the tablet?
• Solution give the control group a tablet with no
  medicine (a placebo). Now subjects in the
  study are blind to their condition
• What is a double blind study?

24
Experiments

• Be careful of potential confounds when designing
  your experiment (esp. in your control group)
• Ask yourself Is everything the same EXCEPT for
  the variable Im manipulating?

25
Experiments

• Strength
• You can attribute differences in the dependent
  variable to the independent variable (causal
  relationship)
• Weakness
• Sometimes too far removed from the real world

26
Who should be in our study?

• Best case scenario random sample from the
  population of interest
• Next best scenario sample that does not differ
  from the population of interest on the dimension
  being studied
• Dont look to us to obtain subjects for you, and
  dont distribute any questionnaires in class

27
How many subjects do we need?

• As many as it takes to detect a difference if
  there really is one
• Depends on the amount of random error and the
  effect size
• See FAQ page for some more guidance

28
Steps for conducting research

• Identify a research question
• Conduct a study/investigation to test the
  research question
• Draw conclusions

29
Steps for conducting research

• Identify a research question
• Conduct a study/investigation to test the
  research question
• Draw conclusions
• You can never PROVE a theory you can only find
  evidence to support a theory or evidence
  consistent with a theory

30

• In class exercise

31
Steps for conducting research

• Identify a research question
• Conduct a study/investigation to test the
  research question
• Draw conclusions
• Share your knowledge (YOUR PRESENTATION!)