-
What is a research question?
- Exploratory research: not much is known
- Asks Is there a relationship between two or more variables?
- Not sure if there is a relationship but think there could be one
- Example:
- “Does age affect what channel people use to get the news?”
- Variables: age and source or news
-
What is a hypothesis?
- Predictive, based on theory
- Predicts how the data will turn out
- Claims there a relationship or difference
- Example:
- “Younger people prefer the Internet as a news source overnewspaper or television”
- Variables
- – Age
- – Source for news
-
Purpose of Hypotheses
- • Provide direction for research
- • Eliminates trial and error
- • Rule out intervening and confounding variables (identify and
- control)
- • Quantification of variables
-
Criteria For Hypothesis
- • Compatible with current knowledge
- • Succinct
- • Testable
- In book:
- Compatible with current knowledge in the area
- Logically consistent
- Stated concisely
- Testable
-
• Says there is no relationship or difference
• The alternative to the hypothesis
• It is always there but rarely stated
Null Hypothesis
-
– Claim found support for hypothesis but really did not
– Reject null hypothesis when is should have been accepted
– not really a relationship or difference but “looks” like there is one
– Claim a result that is not there
Type I Error (Alpha)
-
– Claim there is no support for hypothesis when there really was
– Accept null hypothesis when it should have been rejected
– Claim there is no support for hypothesis when there really was
– Accept null hypothesis when it should have been rejected
Type II Error (Beta)
-
When controlling for Type I error use...
- Significance level – Increasing significance level increases Type II error? (type I maybe?)
- – Set significance level higher
- – in social science use .05 (5% chance finding was in error--
- chance) so go .01 or .001
- – Part of statistical packages
-
When controlling for Type II Error use...
- Power
- • Power analysis indicates how many respondents needed to potentially find an effect
- • Do you have enough respondents to really test the hypothesis?
- • Do you have enough respondents to really test the hypothesis?
- • Lower power increases the possibility of Type II error
- • Acceptable power level is .80
-
Why is significance alone not enough?
- Significance test is important but not everything
- Also need to ask if finding is meaningful
- Part of being meaningful is effect size—how big is the
- relationship or difference
- Significant but small not very impressive or helpful
- Sample size can create significance for small relationships or
- differences
-
In data distribution: scores arranged by magnitude and
frequency
Frequency Distribution
- • Cumulative frequency, add a percent column and cumulative
- percent column
- • Easier to find groups of data
-
Formats of Data Distribution
- Graph: X and Y axis
- Histogram or bar chart
- Frequency polygon: connect points on histogram
-
Frequency distribution is symmetrical and bell shaped
Normal Curve
-
• Shows if scores cluster to the left or to the right
• Many statistics assume a normal curve to data
• May want to standardize data using z-scores
Skewness
-
The two summary statistics
- Central tendencies
- Dispersion
-
The Central Tendencies
- Mode: most common score, can be multiple modes
- Median: midpoint in the data
- – Easier if odd number
- – When even, average two middle scores
- Mean: average of a set of scores, written as M or X with bar
-
Three types of dispersion
- Range
- Variance
- Standard Deviation
-
Dispersion: difference between highest and lowest score
Range
-
Degree to which scores are different from the mean, squared so different unit from the mean
- Variance
- – Small variance all close to mean
- – Large variance wide range of scores
-
same unit as mean, the distance a score is from the mean
– Help to find outliers
Standard Deviation
-
• Extreme scores
• Can affect the mean
• May need to delete from data
Outliers
-
Descriptive Data
- Gives you a general idea of what is going on
- – Is anything unusual happening?
- Mean is a critical statistic
- – Used for other statistical analyses
- Use means to answer key questions:
- – Are variables related to one another?
- – Do the scores differ from one another?
-
The variable being manipulated
Independent Variable
-
The variable being observed
Dependent Variable (think, depends on IV)
-
What are the advantages of experiments?
- 1. Causality
- 2. Control
- 3. Cost
- 4. Replication
-
What reasons for causality give experiments an advantage?
- Only way to get at cause and effect
- Show that A (intervention) caused B (outcome)
- Can control time order of variables
- Can control plausible alternative hypotheses (threats tointernal validity)
-
What control do experiments have that are advantages for experiments?
- Environment (where data is collected)
- Variables (IV and DV)
- Subjects (selection, assignment, and exposure)
-
The reason why cost is an advantage for experiments
It is reasonably low compared to field experiments
-
The reason why replication is an advantage for experiments
Others can repeat the effort to see if they get the same results
-
Disadvantages for experiments
- Artificiality
- Researcher Bias
- Limited Scope
-
The reason why artificiality is a disadvantage for experiments
not how people experience “things”
-
The reason why researcher bias is a disadvantage for experiments
can lead subjects, Use of double-blind to eliminate
-
The reason why a limited scope is a disadvantage for experiments
not all research questions can be examined with experiments
-
What is necessary to prove causality?
- Need to be able to control order
- Design issue
- Significance difference
-
• Cause has to come before the effect (temporal presence)
– Cannot control in surveys
• Has to be a relationship between the variables
– When A changes so does B
– When intervention occurs, behavior changes
• Rule out plausible alternative hypotheses
– Maturation, regression toward mean, history, etc.
Causality
-
What are the 3 categories of experimental design?
- 1. Non-experimental
- 2. Quasi-experimental
- 3. Experimental
-
This category of experimental design is exploratory and rules no plausible alternative hypotheses
X O
Non-experimental
-
What common study is a non-experimental design?
Case Studies
-
This experimental design is a step up from non-experimental but still some plausible alternative hypotheses remain .
Quasi-Experimental
-
What is the most common quasi-experimental group?
Non-equivalent groups
-
Quasi-experimental looks like experimental design but lacks what key ingredient? This determines if it was really an experimental design.
Random Assignment
-
What are the 3 types of experimental design?
- Pretest-Posttest Control Group
- Posttest Only Control Group
- Solomon Four-Group Design
-
What is the key for experimental design?
- If it controls plausible alternative hypotheses through similar groups , or
- Have controls and treatment groups but have similar experiences
-
How do you know control and treatment?
- Control has no X (something still happens)
- Treatment has an X
-
In design notation what is R, O, X, and N?
- R: random assignment
- O: is an observation
- X: is the manipulation/treatment
- N: nonequivalent groups
-
What are the relevant comparisons in a Prestest-Postest Design?
- O1to see if they are the same
- O2 to see if they are different
- O1 and O2 to see if different
- Concern: pretest has an affect
-
What are the relevant comparisons in a Posttest Only Design?
- O1 and O2 for differences
- Concern: groups equivalent
- – Can use other assessments besides your key measures
- – Randomization should take care of it
-
What is a factorial design?
- 1. Includes 2 or more independent variables
- 2. written in number format
- 3. Based in number of variables and levels of the variables
- 4. Can compare different types of treatment
-
How to read a factorial design:
1. 2 x 3
2. 2 x 2 x 2
- Total number is the number of variables
- The number value is the number of levels for the variable
- 1. two variables, variable 1 has two levels and variable 2 has three levels
- 2. three variables, all have two levels
-
More natural experiment that subjects are less conscious of assessment
Field Experiment
-
What are the advantages to field experiments?
- External validity (setting)
- Nonreactive: awareness of assessment
- Good for complex social processes
- Can be inexpensive
- At times may be only option
-
What are the disadvantages to field experiments?
- Some impossible to do in field
- Subject to environmental factors
- So many other variables present, reduces confidence (quasi-
- experimental)
-
What are the advantages of a survey?
- Used to investigate problems in realistic settings
- Cost is reasonable
- Relatively Easy
- Not constrained by geographical boundaries-can be conducted anywhere
- Archived Data--possible to conduct survey strictly without ever developing questionnaire
-
What are the disadvantages of a survey?
- Independent variables can't be manipulated like in lab
- causality is difficult to establish
- Bias results--question can be worded wrong
- Wrong respondents included
- Response rates decline
-
What are the 5 types of survey collection?
- 1. Mail surveys
- 2. Telephone
- 3. Personal Interview
- 4. Group Administration
- 5.Internet Survey
-
Cheapest form of survey collection along with selective sampling, anonymity, answer at own pace, and eliminates interviewer bias. However, it must be self-explanatory, it is the slowest form of data collection, you don't know exactly who is answering and has the lowest response rate
Mail Suvey
-
This form of survey collection has higher response rates than mail surveys but it is limited in questions that can be asked. It is more expensive than mail but less than face-to-face. It is popular in Mass Media Research.
Include the advantages and disadvantages.
- Telephone Survey
- Advantages: Reasonable cost, can include more detailed questions, and faster than mail
- Disadvantages: not really research, no visual demonstration, not everyone is in the directory
-
This form of survey collection is the most flexible, allows for observation and rapport, and harder to terminate once started. However, it is timely and expensive. Interviewer bias is present and the interviews must be scheduled during the weekend or on evenings
Personal Interviews
-
This form of survey collection is a combination of mail surveys and personal interviews. It can take place in a natural setting, but is usually held at a field service location like a hotel ballroom. They can be longer than questionnaire in mail surveys and response rates are high. However, there are suspicions, it is very expensive, and not all survey can use samples that can be tested together in a group.
Group Administration
-
This form of survey collection is low costing, there are no geographic limitations, no specific time restraints, flexible in the approach used to collect data, and can use any type of audio/visual. However, there is no way to ensure that the person recruited is the person answering
Internet
-
What are the general problems with surveys?
1. Subjects or respondents are often unable to recall information about themselves or their activities
- 2. Prestigious Bias – Due to respondents’ feelings of inadequacy or lack of knowledge about a particular topic, they often provide “prestigious” answers rather than
- admit to now knowing something
3. Subjects may purposely deceive researcher by giving incorrect answers to questions
4. Respondents often give elaborate answers to simple questions because they try to figure out the purpose of the study and what the researcher is doing
5. Surveys are often complicated by the inability of respondents to explain their true feelings, perceptions, and beliefs – not because they do not have any but because they cannot put them into words
-
Response rates for survey research
- Mail surveys: 1%-5%
- Telephone surveys: 5%-80%
- Internet Surveys: 5%-80%
- Shopping-center intercept: 5%
- Personal interviews: 40% (depending on recruiting method)
|
|
