COMM 250 Agenda - Week 10

Housekeeping

• C2 - Due Today (Put in Folders)

• RAT 5 – Next Wed.

• RP2 – Nov. 12 (the day before my b-day! :)

Lecture

• Experiments

• ITE 10

Review: Exercise in Coding Open-ended Responses

A Review of Issues with Open-ended Items

Advantages:• Avoids “Framing” an Issue, Eliciting Particular Responses

• Reveals Issues/Repsonses the Researcher Would Have Missed

Disadvantages: • Time Consuming to Code

• Difficult to Categorize Some Responses

Typically Used:• To Get a Preliminary Look at an Issue

• To Ensure Unprompted Responses

Review: The Research Process

Conceptualization• Start with / Develop a Theory and Hypotheses

Planning & Designing Research• Selecting Variables of Interest (IV, DV, Control vars)• Operationalize all Variables (i.e., How to measure the vars?)• Design a Study to Test Hypotheses

Methods for Conducting Research• Plan the Study and Collect the Data

Analyzing & Interpreting Data• Run Statistics and Interpret Results

Re-Conceptualization• Back to the Drawing Board

Experimental ResearchPurpose• To Control Variables (in order)

• To Attribute the Effects to the IV; that is,

• To Infer Causality

Types of Experiments• Pre-Exp. - Typically no Comparison Group

• Quasi-Exp. - IV is manipulated OR Observed, NO Random Assignment of Subjects

• Full Experiments - IV is “manipulated,” Random Assignment of Subjects

Experimental Research (continued)

Experimenters Create Situations . . .• to Control Variables (in order to . . .)• to Attribute Observable Effects to the IV; that is . . .• to Infer Causality

Control by Exposing Subjects to an IV• Manipulating (exposure to) an IV (the “Active Var.”)• Observing (exposure to) an IV (the “Attribute Var.”)

Control by “Ruling Out" Initial Differences• Random Assignment• Pretests

Review: Correlation & Causality

Correlation• Two variables are related (as one varies, the

other varies predictably)

Causation3 “Necessary & Sufficient” Conditions:

• Two variables must be shown to be related

• The IV must precede the DV in Time

• The relationship cannot be due to another “extraneous” variable

Experimental DesignsPre-Experiments (“Pseudo-Experiments”)

1-Group, Posttest Only• Produces a Single Score• E.g.: Exam in School

1-Group, Pretest-Posttest• Produces a Difference Score• E.g.: Evaluation of Corporate Training

Non-Equivalent Groups, Posttest Only• Also Called “Static Group Comparison”• No Random Assignment to Groups• E.g.: Comparing Test Scores for a Training Class to a

Group Who Did Not Take the Training

Experimental DesignsQuasi-Experiments (“Field Experiments”)

1-Group, Time Series Design• Series of Pretests (Baseline) Treatment Series of Posttests

• E.g.: Monitoring the Effects of Blood Pressure Medicine

• Problems: Sensitization, Sleeper Effect, No Comparison Group

Quasi-Equivalent Groups, Pretest-Posttest• Non-Random Assignment to (Treatment, Control) Groups

• Produces a Difference Score

• E.g.: Study of College Classes• Problems: Equivalence (History, etc.)

Quasi-Equivalent Groups, (Multiple) Time Series Design• Combines the Two Designs Above

• Problems: Sensitization, Equivalence, Sleeper Effect

Experimental DesignsFull ExperimentsEquivalent Groups, Pretest-Posttest • Equivalence = Random Assignment of Subjects to Groups• Experiments Provide Control; Reveal Causality (in the Lab) • E.g.: Testing a New Chemotherapy Drug

Equivalent Groups, Posttest Only• Relies on the Random Assignment• Initial Differences COULD Cause Any Observed Effect

• E.g.: Lab Study of New Messaging System

Solomon Four-Group• Combines the Two Designs Above

• Checks for Pretest (Sensitization) Effects• Checks Whether Random Assignment “Worked”

Experimental DesignsFactorial Designs • Multiple IVs (“Factors”); Typically One DV

• Can Be Pre-, Quasi-, or Full Experiments

• Most Common: Quasi- and Full

• Most Common: Posttest Only

Examples –H1: The more competent at comm, the higher income one earns.

2x2 Factorial Design• IVs: Comm Competence (Lo, Hi); Gender (F, M)• DV: Income

3x2x2 Factorial Design• IVs: Competence (L, M, H); Gender (F, M); Occup (BC, WC)

• DV: Income

(Possible) 2 x 2 Factorial DesignHypotheses1. The higher one’s CC, the better liked one is.2. Women are better liked than men.

Independent Variables (IVs)• Comm Competence (“CC”) (measured as Hi / Lo)• Gender (M / F)

Dependent Variable (DV)• Likability Score (could have others)

Control Variable• (Positive/Negative) Attitude

2 x 2 Factorial Design - Example

• IVs: Comm Competence, Gender • DV: Income• Subjects: 20 per cell• Control for: Age, Education, Location

Female Male

Low Comm Competence

20 20

High Comm Competence

20 20

2 x 2 x 2 Factorial Design - Example

• IVs: CC, Gender of Sender, Observer Gender • DV: Income• Subjects: 10 per cell• Control for: Age, Education, Location

Female MaleLow Comm Competence WOMEN 10 10

High Comm Competence WOMEN 10 10

Low Comm Competence MEN 10 10

High Comm Competence MEN 10 10

Experimental Research (Review)

Experimenters Create Situations . . .• to Control Variables (in order to . . .)• to Attribute Observable Effects to the IV; that is . . .• to Infer Causality

Control by Exposing Subjects to an IV• Manipulating (exposure to) an IV (the “Active Var.”)• Observing (exposure to) an IV (the “Attribute Var.”)

Control by “Ruling Out" Initial Differences• Random Assignment• Pretests

In-Class Team Exercise # 10 - Part I:

Design a 3 x 2 Factorial Experiment (draw a Table)

You Must Use These IVs:• Group Size (Use 3 Levels, S, M, L, but choose the # in each)

• Type of Conferencing (Pick 2: Audio, Video, Text, Chat, FtF)

Write out 2 Hypotheses (H1, H2):H1: One Predicting the Effect of Group Size on Group Consensus

H2: One Predicting the Effect of Type of Conferencing on Group Consensus

Declare the DV (You Choose – They Are in Your H1, H2)• E.g., User Satisfaction, Quality of Solution, Time Efficiency

Label the 2 IVs and Label Their Levels

List (at least) 2 Variables you Should “Control for”

Review: Hypotheses

Two-Tailed Hypotheses• Non-directional – researcher predicts a

relationship, but does not specify the nature

• “Comm Competence is related to Annual Income.”

One-Tailed Hypotheses• Directional – researcher predicts both a

relationship AND the direction of it

• “The more Competent one’s Comm, the higher one’s Annual Income.”

Review: Variables of Interest

Independent – influences another variable• IV = “Predictor” variable

Dependent – variable influenced by another• DV = “Outcome” variable

Control – variable one tries to control for• Could “keep constant,” balance across groups,

or extract in the statistical analysis

• Control Var = “Concomitant” variable

Extraneous Variables

Intervening Var – explains relation bet IV, DV

• “The a Person’s Comm Competence (CC) (the IV), the the Salary (the DV).”

• Since Competence, per se, doesn’t get you $, “Job Function” is an Intervening Var.

Extraneous Variables (continued)

Confounding Var – obscure effects• “Surpressor” Var. reduces the effect of an IV• CC could # of Friends, but also difficulty of

chosen job, which in turn time for friends.

• “Reinforcer” Var. increases the effect of an IV• CC could # of Friends, but also # of events one

attends, which in turn would further # of friends.

Lurking Var – explains both IV and DV• Perhaps the var “Extroversion” affects both CC

and # of Friends.

Statistics

Descriptive Statistics: a way to summarize data

Inferential Statistics: strategies for estimating population

characteristics from data gathered on a sample

Descriptive Statistics

Measures of Central Tendency Used to describe similarities among scores What number best describes the entire

distribution?

Measures of Dispersion Used to describe differences among scores How much do scores vary?

Descriptive Statistics

Measures of Central Tendency

Mean The Average Medium The Middle Score Mode The Most Common Score

Measures of Dispersion• Range

The Highest & Lowest Scores

Variance A Measure Of Dispersion Equal To The

Average Distance Of The Scores, Squared, From The Mean Of All Scores, Divided By N

Standard Deviation The Square Root Of The Variance

(Dispersion About The Mean, Based In The Original Units)