traits or characteristics of people, such as age or gender

can't be  manipulated

whenever exposure to an independent variable is not manipulated by researchers themselves

less control over the experiement

Random Assignment

(randomization)

difference score

(gain score)

occur when different experimenters consistently administer particular manipulations of the independent variable

it wont be clear if participants are responding to the differences in treatments or the differences in the way that they are administered

(A kind of confoudning variable)

A variable that causes both the independent and dependent variable "An unpleasant surprised when discovered"

Control variable

(concomitant)

Matched-pairs design

(participant or subject matching)

Full expierment

Quasi-Experiment

Preexperiment

demonstrate the highest degree of control because the independent variable is manipulated by the researcher and research participants are randomly assigned to create two or more equivalent conditions.

Must have:

Control of independent variable manipulation

two or more conditions

random assignment

Either manipulate or observe the independent variable and may have one or more conditions.

No random assigment when more than one condition

Created using pretests to assess whether there are some important intital differences between the conditions

Manipulate or observe the independent variable and more have one or more conditions.

No assigment on multiple conditions

No pretests used to find intial differences

One-group posttest only design

Type of preexperimental design

a single treatment group is exposed to the independent variable and then assessed on a posttest

One-group pretest-postest design

(before-after designs)

Preexperimental design

A single treatment group is given a pretest then exposed to the independent variable then given a posttest

Posttest-Only Nonequivalent groups design

(static group comparision design)

Preexperimental design

nonrandomly assigns research participants to a treatment or a control group and then measures them on a posttest.

Single-Group Interrupted Time Series Design

(time series design)

(Quasi-Experimental Designs)

involves giving a series of pretests to a single group prior to an expiremental treatment followed by a series of posttests

(Quasi-experimental design)

nonrandomly assigns research participants to a treatment or control condition, measures them on a pretest, exposes one group but not another to the treatment, and then measures both again in a posttest

(quasi-experimental design)

nonrandomly assigning participants to treatment or control group and measuring them on a series of pretests and posttests

(Full expirement design)

randomly assigns participants to treatmen or control group and administers a pretest and posttest

(full experiment design)

Participants randomly assigned to treatment or control group and given a posttest

(full experiment design)

Applied survey research conducted to measure public opinion. Includes:

Evalutation research

market research

political polls

type of evaluation research used to idenitfy structures in social systems based on the relations among a system's components

secondary data analysis

a list of the population from which they will sample

*Ideally it will list all the members of the population

the group that gets nothing at all-baseline

BUT still answer the same questionnaire

Pre-experimental designs

Single cross-sectional survey or one-shot case study

NR pre-program         Program              post-program

(No pretest)       >Treatment>      Post-test/questionnaire

X                          O1

Pre-experimental design

One group pretest-posttest design

Quasi-experiments

Pretest/posttest quasi-equivalent group design

NR            O1          X          O2

NR              O3                    O4

·      Problem: Groups not equivalent at outset and differences may account for posttest differences; harder to control for rival explanations; weaker internal validity

·      Very often quasi-experimental designs are only viable evaluation strategies in PR/AD

Random assignment of subjects to treatment and control groups

·      Equal opportunity to be assigned to experiment

·      Control=groups get placebo or no treatment

Pre and post-treatment measures permit powerful cause-effect inference

·      Note: lab experiments weak in external validity

Full experiment

Pretest-posttest control group design

R         O1           X          O2

R          O3                      O4

·      Threats : history, testing effect, sensitizing interaction of pretest, maturation, like effect (X>control group)

Full Experiment

Posttest Only Control Group

R           X         O1

R                      O2

·      Threats: no direct measure of prior knowledge, maturation, history, like effect; effects may be under-estimated

Factorial Design

·      More than one Independent variable

·      Levels – Number of options within the variable

·      Example: race & Gender 2x2 factorial design

o   Numbers represent number of levels that each independent variable has

o   Number of numbers represents # of Ind. Variables

§  Race- African- American & Caucasian

§  Gender- male & and female

·      Can only draw fact design diagram if there are 2 or more IVs

o   Cells are alwaysIVS                               

·      Race along one axis, gender along other

o  I: African American Male

o   II: African American Female

o  III: Caucasian Male

o   IV: Caucasian Female

Square drawing with roman numerals

·      Researcher not actually measuring things- the respondent sends in observations and reports

·      tendency to please the surveyor

Political Polls

Evaluation Research

Market Research

o   Formative: before

§  creates baseline to compare from

o   Summative: after

§  Did what you performed work?

o   Need analysis: what would you like?

o   Audit & organize feedback survey

Evaluating product & levels of consumption

o   Readership survey

§  Media outlets use to evaluate

§  Audience ratings-

·      Trend- two or more points in time

·      Cohort- specific subgroups over time

o   not necessarily the same exact people

·      Panel- same people over time

o   Panel attrition- dropouts

·      Each have the issue of people dropping out

·      Only calculated with probability samples

·      Paper & pencil strives for 60%

·      Way you deploy survey can influence rate

·      How many did you send to?

·      How many were valid (not returned)?

·      How many were completed? (y)

o   Sent-returned=X

o   Y/X= response rate

·      Only calculated with probability samples

·      Paper & pencil strives for 60%

·      Way you deploy survey can influence rate

·      How many did you send to?

·      How many were valid (not returned)?

·      How many were completed? (y)

o   Sent-returned=X

o   Y/X= response rate

·      Ask only one thing…. Friends or family? (treat groups differently)

o   What you are asking about? Don’t confuse respondent

·      Don't lead respondent to answer

o   “Smoking pot isn’t good, is it?” already saying what you feel about it, and ppl want to Please the surveyor

·      Approach concepts from several angles

o   asking several different questions about the same thing

·      Closed-ended or open-ended?

o   Open-ended- can find responses you normally wouldn’t find- harder to compile yet will allow for emerging concepts

·      Does order matter?

o   Ease into tough questions

·      a group or class of subjects, variables, concepts or phenomena.

o   Census- looking at all in population

·      subset of population that is representative of entire population

o   Brings limitations in research

·      follows mathematical guidelines

o   simple random, etc. every person in the population has a chance to be a part of the survey

·      - does not follow mathematical guidelines

o   convience, volunteer- must acknowledge that some parts of pop may be left out (non college students, etc)

Convenience

Volunteer

Purposive

Snowball

·      readily accessible

o   access a certain group of people because they are convenient to you

·      asking for people to participate- ppl willing to participate

o   not random, b/c people are self-selecting= certain amount of bias, have some incentive

·      includes subjects based on characteristics

o   specific item narrows down the sample- major, gender, etc.

·      ask for referrals- after ppl participate, then ask for them to refer others to participate

o   number table

o   Random number generator

o   RDD- random digit dialing

o   Stratified Random sampling- looking through layers of random sampling to establish a certain number/aspect of a person- narrowing a group down based on a certain criteria- at beginning everyone has an equal chance to be included