Term
|
Definition
• determines all we can and cannot (45cfr46= federal regulation)
• once we have ethical study we then turn to accuracy (validity) |
|
|
Term
|
Definition
• concerned with how accurately you’ve measured abstract constructs
o look at validity and reliability
• good research requires good measures |
|
|
Term
|
Definition
• concerned with how accurately your sample can speak for the population
• to adequately describe a population, you need a good (representative) sample |
|
|
Term
|
Definition
• concerned with the accuracy of your conclusion that X causes/does not cause Y
o must meet the THREE CRITERIA FOR CAUSALITY
• good (causal) research will adequately meet all three |
|
|
Term
| what type of study has a stronger internal validity but tend to have a weaker external validity (application to real world)? |
|
Definition
| true experiments conducted in a lab |
|
|
Term
| what type of study had a stronger external validity with a weaker internal validity(control for z-factors)? |
|
Definition
| field research done out in real-world |
|
|
Term
| where did Dr. Exum go to college? |
|
Definition
|
|
Term
| why do two studies on the same topic differ? |
|
Definition
• Measure of X & Y may be different
• Samples (and sample sizes) may be different
• Research design may be different
o May not have established group equivalence
o May have different control variables |
|
|
Term
what are the three ethical principles identified in the Belmont Report?
*began 1974 & finished 1979 |
|
Definition
respect for persons
beneficence
justice |
|
|
Term
| what is "respect for persons"? (Belmont Report) |
|
Definition
o Should view individuals as autonomous (self-governing)
• Cannot force people to participate |
|
|
Term
| what is an “Informed Consent Form”? (result of "respect for persons") |
|
Definition
Describes study
• Indicates participation is voluntary
• Participant & Researcher must sign |
|
|
Term
|
Definition
• determines all we can and cannot (45cfr46= federal regulation)
• once we have ethical study we then turn to accuracy (validity) |
|
|
Term
|
Definition
• concerned with how accurately you’ve measured abstract constructs
o look at validity and reliability
• good research requires good measures |
|
|
Term
|
Definition
• concerned with how accurately your sample can speak for the population
• to adequately describe a population, you need a good (representative) sample |
|
|
Term
|
Definition
• concerned with the accuracy of your conclusion that X causes/does not cause Y
o must meet the THREE CRITERIA FOR CAUSALITY
• good (causal) research will adequately meet all three |
|
|
Term
| what type of study has a stronger internal validity but tend to have a weaker external validity (application to real world)? |
|
Definition
| true experiments conducted in a lab |
|
|
Term
| what type of study had a stronger external validity with a weaker internal validity(control for z-factors)? |
|
Definition
| field research done out in real-world |
|
|
Term
| What is "beneficence"? (Belmont Report) |
|
Definition
o Do no harm
o Maximize benefits & minimize harms
• Physical, social (status, standing), psychological, economic, legal, etc. |
|
|
Term
| what is a risk/benefit assessment? ("beneficence") |
|
Definition
| • Harms must be justifiable |
|
|
Term
| what is "justice"? (Belmont Report) |
|
Definition
o Fair sharing of the benefits and burdens of research
• Must justify why you are focusing on only one group |
|
|
Term
| what is an "equable process for selecting research patients"? ("justice") |
|
Definition
| • If study focuses exclusively on certain groups, you must justify why |
|
|
Term
|
Definition
| evolving document of specific rules that researchers must follow to receive federal funding |
|
|
Term
| What are "IRBs"? (Institutional review boards) |
|
Definition
• Institutions governed by the CFR are required to have one of these to oversee “human subjects research”
Minimum of 5 people, diverse backgrounds
• Reviews research proposals and decides if the study complies with 45 CFR 46
Study must have their approval before starting |
|
|
Term
| What is the purpose of an "IRB"? |
|
Definition
reviews proposals to:
• Determine if the risks of the study acceptable
• Determine if there are adequate safeguards in place for participants |
|
|
Term
|
Definition
ABSTRACT
o Abstract-> Concrete
o Construct (concept)
• Abstract idea
Problem drinker; chronic offender; poverty
We all have a general idea of what these things mean, but our specific definitions may vary |
|
|
Term
|
Definition
CONCRETE
o Concrete->Abstract
o E.G.- Men in class have been in more fights than have the women…induce that men fight more than women
o Indicator
• The concrete (specific) way we measure a concept
Aka- Our “measure” or “operational definition”
You travel from CONCEPT to INDICATOR through deduction. |
|
|
Term
|
Definition
ABSTRACT
• “A general explanation for how things work or how they come to be”
o poverty cause crime
o juvenile delinquency is the result of poor parenting
o X-> casual chain
• Built from constructs |
|
|
Term
|
Definition
o Statement
o Deduced from theory
o Predicts a relationship b/t two or more variables |
|
|
Term
| what is an"operational definition"? |
|
Definition
|
|
Term
|
Definition
(concept)
• Abstract idea
Problem drinker; chronic offender; poverty
We all have a general idea of what these things mean, but our specific definitions may vary |
|
|
Term
|
Definition
X, Y, & Z
• Measure of a concept (e.g.- indicator) that has at least 2 values (or “scores” or “ attributes”)
• P. 51
• Examples:
o Variable-> Attribute
• Sex->male, female
• Race-> white, black, etc |
|
|
Term
|
Definition
• A measure of much variability is in the set of scores for an indicator
• “movement” |
|
|
Term
|
Definition
may be numbers
• This does not necessarily mean they are quantitative
o May be numeric codes for qualitative data
can be:
• Qualities that cannot be ranked (no higher/lower)
o Northeast, South, Midwest, West
• Qualities then can be ranked (higher/lower)
o Lower class, middle class, upper class
Quantities that are precise amounts
• 0 times arrested, 1 time, 2 times, 3 times, etc.. |
|
|
Term
| tells us something about the type of attributes for a given measure |
|
Definition
|
|
Term
|
Definition
• A measure with no variance
o All scores are the same |
|
|
Term
|
Definition
• “As X moves does Y move in some general direction?”
o If yes (even just a little) then there is a relationship
o They can move in the same direction or opposite direction….either way there is a relationship |
|
|
Term
|
Definition
• If two variables (X & Y) are causally related, then X is the originator of Y
o X causes Y
X -> Y |
|
|
Term
|
Definition
o Must have a correlation
• Scores for X & Y move together
o Must have the proper temporal order
• X occurred before Y
o Must rule out rival explanations (or “spuriousness”)
• Rule out the possibility that X & Y are related solely because of some other variable (Z) |
|
|
Term
|
Definition
not causal and attributed to some other factor we’ll call Z
Z
/\
X Y |
|
|
Term
|
Definition
|
|
Term
|
Definition
|
|
Term
|
Definition
"z factors"
• Are rival explanations for why X & Y are correlated
o Rival to the idea that X causes Y
• We must eliminate (or “control for”) these explanations |
|
|
Term
|
Definition
represents descriptive features
o E.g.- race; the type of crime committed; narrative accounts/stories
o Sometimes coded with numbers
• Still qualitative; just with numeric codes |
|
|
Term
|
Definition
represents how much of a construct
o E.g.- age, number of prior arrests |
|
|
Term
|
Definition
| o Use data to describe our sample |
|
|
Term
|
Definition
o Use data to describe our population
o Requires “p-values” (e.g. p<.05) |
|
|
Term
|
Definition
• Studying a phenomenon at one point in time
o E.g.- a one-time survey given to participants
• Provides a snap shot at that moment |
|
|
Term
| what are "cross-sectional studies" NOT good at? |
|
Definition
examining causal chains
• Hard to show that X comes before Y in cross sectional data
• To show temporal order, you could try retrospective measures
o Ask participants to think back about something in past (e.g.-depression level at beginning of the year)
• But can people accurately remember the past??? QUESTIONABLE. |
|
|
Term
|
Definition
• Identify yourself as researcher
• Observe/interview
o but do not participate in the behavior
o thus, may not have a complete appreciation or understanding of the behavior
• Important to build rapport/trust
o Helps you become “invisible”
• Problem: Reactive Effects |
|
|
Term
|
Definition
• Identify yourself, but also participate in the behavior (to some degree)
o Gives you firsthand experience/insight
o Reactive effects still possible
• Less? You gain trust by going through what they go through. |
|
|
Term
|
Definition
• Do not identify yourself
• Participate in the group’s activities as a “member”
• Reactive effects? Should be virtually zero.
• Measurement issues
o Asking a lot of questions may blow your cover
o May lose your objectivity (‘going native’) |
|
|
Term
|
Definition
| o people don’t behave naturally if they know they are being watched |
|
|
Term
|
Definition
Qualitative
-the simplest or “lowest” level of measurement
- non-hierarchical categories
• no “greater than/lesser than”
• Common CJ examples:
o Sex (m/f)
o Race (w, b, o)
o Marital status
o Narratives (responses to questions)
-not numerically meaningful
• they do not measure quantity
-what you do with these data?
• Count (or determine the percentage) in each group |
|
|
Term
|
Definition
qualitative
-attributes represent hierarchical categories
• do reflect “greater than/lesser than”
• but do not convey PRECISE quantities
o Common CJ examples:
• Level of agreement (sd, d, a, sa)
• Level of frequency (never, rarely, often, always)
• Prison security (min, med, max, super-max)
-because the data does not measure a precise amount, you cannot compute an average
-what can you do with these data?
• Count (or determine percentage) in each group
• Can also rank order the attributes |
|
|
Term
|
Definition
-quantitative
-the most precise or “highest” level of measurement
-attributes reflect an actual, precise quantity
• Common CJ examples:
o Age
o Grade
o Weight
-what can you do with these data?
• Count (or determine percentage) in each group
• Can also rank order the attributes
• Higher order mathematical operations, like the mean
-To be a ratio level of measurement:
• All the attributes must correspond to a single specific value |
|
|
Term
|
Definition
multiple-item indicator
combine the indicators to create a "scale score" |
|
|
Term
| forumla to calculate crime rate per 100,000 people |
|
Definition
| (#crimes/population)*100,000 |
|
|
Term
|
Definition
o AKA: Frequency Table
• Nominal, Ordinal, & Ratio |
|
|
Term
|
Definition
nominal & ordinal
• Bars don’t touch (discrete categories) |
|
|
Term
|
Definition
ratio
• Bars touch (fluid, continuous data) |
|
|
Term
|
Definition
(bell shaped)
• one peak with two tails
• right side is mirror image of left |
|
|
Term
|
Definition
• the skew is always in the tail
EX- Left skewed: will skew in left tail |
|
|
Term
|
Definition
• two peaks (sets of clusters)
• 3 peaks= tri-modal (and so on) |
|
|
Term
|
Definition
flat
• roughly same number for every attribute |
|
|
Term
| three measures of central tendency |
|
Definition
• numbers that will tell us where a variable’s attributes TEND to fall
mean
median
mode |
|
|
Term
|
Definition
• The average
• “the value around which all deviations sum to zero”
• great measure; use it when you can
• designed for ratio level measures
• highly influenced by skewness and outliers
still mathematically correct
but can now be misleading |
|
|
Term
|
Definition
• the attribute falling in the middle of a rank ordered set of scores
• attribute that falls at the 50th percentile
• not as functional as the mean
• Not sensitive to outliers/skewness
• The variable must be ordinal or ratio (must put attributes in rank order; you cant rank nominal) |
|
|
Term
|
Definition
• simplest measure of central tendency
• the attribute that occurs most often (has the highest frequency)
NOT how many times it occurs
• Not as functional as the median (or mean)
• A variable can have more than 1 (bi-modal)
• Not sensitive to outliers/skewness
• The only central tendency measure you can use with nominal data |
|
|
Term
| Rules for Using Measures of Central Tendency |
|
Definition
• 1) Use the mean whenever it is appropriate
o ratio data that is normally distributed (or “approximately normal”)
• 2) If you cant use the mean, use the median
o ordinal data, or ratio data that are highly skewed/outliers
• 3) If you cant use the median, use the mode
o when you have nominal level data |
|
|
Term
|
Definition
measure the amount of “movement”
• range
• standard deviation
• variance
-all are designed for RATIO measures |
|
|
Term
|
Definition
• simplest measure
• range=highest-lowest
• lower range=more variability
• Weaknesses of the range:
o Sensitive to outliers
o Ignores the variability of the scores “in the middle” |
|
|
Term
|
Definition
uses all the scores (not just the extremes
o -remember “deviations”
o - A SPECIAL kind of average
o -the “average” deviation from the mean
-tells us how far scores can move-on “average”-around the mean
-Larger SD’s=more spread; “fatter” distribution |
|
|
Term
| How do you compute the standard deviation? |
|
Definition
Compute the deviations from the mean
o -2, -1, 1, 2
o -Square them
o 4, 1, 1, 4
o -Average these squared deviations(VARIANCE)
o (4+1+1+4)/ 4=2.5
o -Take the square root of this average
-square root of 2.5= 1.6 drinks
-this is your SD! |
|
|
Term
|
Definition
think of it as the TOTAL amount of movement
o -mathematically: variance= SD 2(squared)
o -Larger variance=more variability
The special relationship between the Mean, SD, and the Normal Curve |
|
|
Term
| The Pearson r & Venn diagram are considered what type of relationship? |
|
Definition
|
|
Term
| in the Pearson r how do you interpret the sign & the number? |
|
Definition
{-} negative relationship
{__} positive relationship |
|
|
Term
| What constitutes a weak relationship? |
|
Definition
|
|
Term
| What constitutes a moderate relationship? |
|
Definition
|
|
Term
| What constitutes a strong relationship? |
|
Definition
|
|
Term
| What does the Explained Variance tell us? |
|
Definition
| % of the variance in Y that is attributed to X |
|
|
Term
| How do you calculate the Explained Variance from the Peason R? |
|
Definition
|
|
Term
| How do you rule out spuriousness in non-experimental studies? |
|
Definition
-try to eliminate Z through statistical techniques
1. Allow X to happen naturally and then measure it
2. Measure Y
3. Measure the Z factors that you think might render your XY correlation spurious
4. Multivariate statistical techniques to see if X and Y are correlated above and beyond the influence of Z |
|
|
Term
| How do you rule out spuriousness in experimental studies? |
|
Definition
-rule out spuriousness Methodically
-Try to establish group equivalence |
|
|
Term
| What are "True Experiments"? |
|
Definition
Pretest-Post Test experimental design
Post Test only experimental design
Factorial experimental design |
|
|
Term
| How does the Pretest-Post test experimental design work? |
|
Definition
| pretest measured on dependent variable, applied stimulus then re-measure (post test) |
|
|
Term
| How does the Post test only experimental design work? |
|
Definition
no pretest done
can reduce the possibility of the test being a threat to validity
key is randomization |
|
|
Term
| What are the Quasi-experiments? |
|
Definition
Non-equivalent control group design
cohort design
time series design |
|
|
Term
| non-equivalent control group design |
|
Definition
• Typically uses matching
• Identifying a pair of participants who are “identical” on a variable you want to control for.
• Assign one to treatment and other to control |
|
|
Term
|
Definition
• Does not use matching
• Treatment and control group are dif. Cohorts
• Typically run in succession (dif. Groups of participants at dif. Points in time) |
|
|
Term
|
Definition
• Same group of people over time.
• Only one group of participants
• Participants serve as own control group |
|
|
Term
| Why is matching inferior to random assignment? |
|
Definition
it only controls for those variables on which you match
random assignment: controls for all possible variables (in theory) |
|
|
Term
| Why are True Experiments stronger research designs than Quasi-experiments? |
|
Definition
True experiments use random assignment
& they are better able to meet the 3 criteria for causality |
|
|
Term
| what are the threats to internal validity in non-experimental studies that we discussed in class? |
|
Definition
o 1. Incorrect Temporal Order-
• Correlation studies often measure X and Y at the same time (cross-sectional study)
• Can be difficult to determine which happens first
o (depression→ low gpa)
o (low gpa→ depression)
2. Omitted Variable Bias
-Occurs when you fail to control for relevant z factors
-the variables “omitted” from your analysis b/c you forgot to (or could not) measure it/control for it. |
|
|
Term
| Selection Bias (Threat #1 to internal validity in experimental studies) |
|
Definition
treatment & control group offer some important factor at the start of the story
-Ex: more men selected than women
-less of a problem when you use random assignment |
|
|
Term
| Experimental Mortality (Threat #2 to internal validity in experimental studies) |
|
Definition
Aka: differential attrition
a potential problem in longitudinal studies
if it occurs equally across treatment & control group then the problem cancels out
*differential attrition is a more serious problem |
|
|
Term
| Maturation Effects (Threat #3 to internal validity in experimental studies) |
|
Definition
maturation=changes in behavior that occurs naturally within the person over time
-if occurs differentially then it's a big problem
Ex: 2 groups; same crime; 1 pays a fine & other gets 10yrs in prison |
|
|
Term
| Statistical Regression (Threat #4 to internal validity in experimental studies) |
|
Definition
Aka- regression to the mean
this is the natural tendency for behavior to ebb-&-flow around a mean
-a bit like maturation but here change is a cyclical (some days up; others down) |
|
|
Term
| Diffusion (or "contagion") Effects (Threat #5 to internal validity in experimental studies) |
|
Definition
sometimes treatment "spills over" into the control group
-so X is given to both groups
-no longer have a true counterfactual |
|
|
Term
| Compensatory Rivalry (Threat #6 to internal validity in experimental studies) |
|
Definition
| the control group realizes they are not getting "X", so they change their behavior |
|
|
Term
| Hawthorne Effects (Threat #7 to internal validity in experimental studies) |
|
Definition
| the treatment group knows they are getting "X", so they change their behavior |
|
|
Term
| How do you prevent the threats of Compensatory Rivalry (#6) & Hawthorne Effects (#7)? |
|
Definition
If possible, "blind" (or "mask") participants to their condition
Don't let them know if they are in the treatment or control group
-may be difficult to do |
|
|
Term
| Percentage vs. Valid Percentage |
|
Definition
valid percentage more useful b/c it's based on people who answered the entire survey
-more accurate |
|
|
Term
| Why would you NOT use a True Experiment? |
|
Definition
| ethical and/or practical reasons |
|
|
Term
|
Definition
row percentage: if you add all the rows= 100%
column % downwards
use when X & Y are nominal or ordinal |
|
|
Term
| special relationship between the mean, SD, & normal curve |
|
Definition
mean +/- SD captures 68% of scores under curve
mean= 2SD= 95%
mean= 3SD= 99% |
|
|
Term
|
Definition
for 2 ratio level variables (bi-variate)
Aka- Pearson product moment correlation coefficient
# ranging from 0 to l1l |
|
|
Term
| Spurious relationship (Venn diagram) |
|
Definition
| if the circle for Z completely overlaps the relationship between X & Y then Z can account for the relationship between X & Y it is spurious |
|
|
Term
|
Definition
| the probability (likelihood) if Y & X had not occured |
|
|
Term
| how do you determine if there is a Bivariate relationship? |
|
Definition
| As X moves Y also moves in some general pattern |
|
|
Term
|
Definition
a graph of each X/Y pairing
-independent variable X on the x-axis
-dependent variable Y on the Y axis
may include the "line of best fit"
-the straight line that is as close as possible to all data points (**minimizes wiggle) |
|
|
Term
|
Definition
controls for all possible variables (in theory)
According to probability theory you can pick up all Z factors (in theory) |
|
|
Term
|
Definition
controls for only those variables on which you match
***inferior to random assignment
o Identifying a pair of participants who are “identical” on a variable you want to control for
o Assign one to treatment and other to control group
PROBLEM: difficult to match on a lot of variables |
|
|
Term
|
Definition
• occurs when you fail to control for a relevant Z factor
o the variable is “omitted” from your analysis because you forgot to (or could not) measure it/control for it
• you cant possibly measure/include every possible Z factor, but you should try to control for the factors that are most likely to be correlated with X & Y
• If at all possible, control for a “past” measure of Y |
|
|
Term
|
Definition
| -concerned with how accurately our sample can speak for the population |
|
|
Term
|
Definition
-the entire collection of “elements” (people, places, or things) we are interested in describing.
-Studying populations can be difficult/costly |
|
|
Term
|
Definition
| -subset of the population |
|
|
Term
|
Definition
• Uses random selection (not random assignment)
o Selects people at random
• Everyone has an equal and independent chance of being selected
• As a result, our sample should look a lot like our population
• We will know the probability of an element being selected into the sample
we can only estimate sampling error (margin of error) in these |
|
|
Term
|
Definition
• No random selection
• We do not know the probability of being selected
• Typically easier and less expensive to create
o Very common, despite their problems |
|
|
Term
| 3 types of non-probability samples |
|
Definition
o Convenience sample
• Sometimes called “Reliance on Available Subjects” (p. 155)
o Quota sample
o Snowball sample |
|
|
Term
| steps to getting a random sample |
|
Definition
• get a sampling frame
o master list of everyone in the population
• assign each element an ID #
• determine the number of digits in the largest ID#
o (ex: 35= 2 digits, 522= 3 digits, 1,332= 4 digits)
• select a starting place on the table
• read the appropriate number of digits along right hand side
• skip repeated numbers or numbers that don’t match an ID# |
|
|
Term
|
Definition
| -a statement that predicts a relationship between two variables |
|
|
Term
| If p<____ then our relationship is “statistically significant” (i.e.- it is real) |
|
Definition
.05
less than 5% chance
o P=the chance of getting your r assuming the null hypothesis is true
o Or, think of it as the chance that your r is not “real” |
|
|
Term
|
Definition
probability value
produced from: pearson r, t-test, chi-square, regression |
|
|
Term
|
Definition
• H0: r=0
o Fisher’s hypothesis
• FYI: Your hypothesis is the “alternative hypothesis” or the “research hypothesis” |
|
|
Term
| if a finding IS statistically significant we ______ the null |
|
Definition
|
|
Term
| greater than/equal to 5%, |
|
Definition
then we fail to reject the null
o Assume there is no relationship in the population |
|
|
Term
|
Definition
this is the cut-off point we use to determine statistical significance
• Typically, alpha= .05, but it doesn’t have to be
• Some make alpha= .01, so now sample findings must fall outside +/- 3 SE to be significant |
|
|
Term
|
Definition
How good/accurate are our measures (operational definitions) of our constructs?-
Concerned with a measure’s
• Reliability
• Validity |
|
|
Term
|
Definition
the indicator
construct+error |
|
|
Term
|
Definition
| -means “consistency” or “repeatability” |
|
|
Term
|
Definition
o AKA: Internal Consistency or Scale Reliability
o Performed on scales (not single-item indicators)
o Looks at how well the indicator scores “hang together” (ex- how well they correlate together)
• If all the indicators are measuring the same construct, then they should all have similar scores |
|
|
Term
|
Definition
• A statistic that measures internal consistency
• Values range from 0.00 to 1.00 (always positive)
• Higher values=greater consistency
• Rule of thumb…
o “Good” scales should have a minimum of 0.70
o Preferably 0.80 and higher |
|
|
Term
|
Definition
• Used with single-item indictors or scales
• If your measurement instrument is reliable, then you should get similar scores each time you administer I to a particular person
o EX- bathroom scale |
|
|
Term
| rule of thumb for strong correlation in test-retest reliability |
|
Definition
|
|
Term
|
Definition
means accuracy
• does your measure accurately capture what you think it is measuring? |
|
|
Term
different types of validity?
-all of which can be used with single-tem indicators or scales |
|
Definition
• Face Validity
• Content Validity
• Criterion Validity
• Construct Validity |
|
|
Term
|
Definition
• no math involved
• just by looking at the item (“on its face”), does it appear to measure what you want it to measure |
|
|
Term
|
Definition
requires math
• Example:
o is a breathalyzer valid?
• Take blood samples from drinkers (criterion)
• Take breath readings from same set of drinkers
• Correlate the two sets of scores
• Strong positive correlations=strong criterion validity
• Rule of thumb:
o Minimum of +0.70
o Prefer +0.80 and higher
• Problem:
o Sometimes hard to find the gold standard measure of our construct
• Low self control? Neighborhood disorder? Social bonds?
o So, may not be able to examine this |
|
|
Term
|
Definition
• requires math
• concerned with how well your indicator correlates with other theoretically-related variables
• Correlations b/t your measure and the “other” measure may be positive or negative
• It depends on the theoretical relationship
• The scores should be modestly correlated
o Not too weak (they should be correlated)
o Not too strong (otherwise, your indicator may actually be a measure of the “other” construct)
• Rule of Thumb:
o Between 0.25-0.60 |
|
|
Term
|
Definition
|
|
Term
| Guideline #1 for writing good survey questions |
|
Definition
| • Consult the literature for pre-existing questions |
|
|
Term
| Guideline #2 for writing good survey questions |
|
Definition
• Use open-ended questions sparingly
o Participants don’t like to write answers (missing data; they will avoid filling out open questions) |
|
|
Term
| Guideline #3 for writing good survey questions |
|
Definition
• With Likert scales, decide if you want/need a “Neutral” option
o I approve of the use of the death penalty
• SD
• D
• N
• A
• SA
o “neutral” encourage “fence sitters”
o But sometimes you may feel “neutral” is needed |
|
|
Term
| Guideline #4 for writing good survey questions |
|
Definition
• Write questions at a low reading level
o Use short, simple sentence structure, simple words, etc.
• 40% of US population reads at a 6th grade level |
|
|
Term
| Guideline #5 for writing good survey questions |
|
Definition
• Avoid negatively worded “stems” if possible
o Can add confusion (error!)
o I believe juveniles should not be tried as adults
• SD, D, A, SA
o Better question: I believe juveniles should be tried as adults. |
|
|
Term
| Guideline #6 for writing good survey questions |
|
Definition
• Avoid double negatives
o Confusing (error!)
o Is it not unlike you to call the police if you witnessed a crime?
• Yes
• No |
|
|
Term
| Guideline #7 for writing good survey questions |
|
Definition
• Avoid double barreled questions
o These are two questions within one
o I believe the death penalty is cruel and unusual punishment and should not be used under any circumstances
• SD, D, A, SA
o Introduces error!
• Better Approach: Split them into separate questions |
|
|
Term
| Guideline #8 for writing good survey questions |
|
Definition
• Make sure your response options are exhaustive (cover all possible answers)
o EX. Poor question:
• How many times have you received a speeding ticket?
1-2
3-5
6-10
• There is no response if you have never received a ticket
• ERROR!
• An “Other:_____” option can help to make a response set exhaustive. |
|
|
Term
| Guideline #9 for writing good survey questions |
|
Definition
• Make sure your response options are mutually exclusive (no overlap)
o EX- poor question:
• How many times have you received a speeding ticket?
0
1-2
2-5
5-10
10+
• Error! |
|
|
Term
| Guideline #10 for writing good survey questions |
|
Definition
• make sure your questions and answers make sense
o EX- poor question:
• Occasionally, I worry about being a crime victim.
Never true of me
Rarely true of me
Sometimes true of me
Often true of me
Always true of me |
|
|
Term
|
Definition
-some people selected to be in your sample won’t participate
Rate=(# participants/# in sample)*100= ____% |
|
|
Term
| rule of thumb for acceptable response rate |
|
Definition
o Minimum response rate= 50%
• General for social science is only 30%
o Good response rate= 60%
o Great response rate= 70+% |
|
|
Term
| Who is less likely to participate in research studies? |
|
Definition
• Men
• Nonwhites
• Young
• Less educated (lower SES/social economic standing) |
|
|
Term
|
Definition
|
|
Term
| Dillman's recommendations on how to generate a high response rate |
|
Definition
send out "tickle letter" to pique interest
make sure all materials appear professional
make it personal
o Use participant’s name in address and greeting
o Hand sign cover letters
o Use stamps rather than metered postage
o Place stamps slightly askew
make it user-friendly
o Easy to red font, lots of white spaces
o Should be a short survey; at most around 8-10pgs
o Lots of close-ended questions (if possible)
o Include a self-addressed stamped envelope for the survey to be returned
• Your opening questions should be easy to answer, non-offensive, and relevant to the purpose of the study ("hook")
• Incentivize participants
o If given in advance, may spark the “norm of reciprocity”
• Incentives need not be expensive
• Send out reminder postcards approximately 2 weeks after surveys
• If still no response, 2 weeks later resend survey again |
|
|
