Lecture Content Flashcards
What are single case designs and what is the main issue with them?
a design where the IV is assessed using data from a single participant
all single case designs lack generalizability! (due to there only being one participant)
What are the types of single case designs?
- ABA Reversal Design
- behaviour observed at baseline –> treatment –> baseline - ABAB Reversal Design
- behaviour observed at baseline –> treatment –> baseline –> treatment - Multiple Baseline Design
- Observing behaviour after manipulation is introduced under several circumstances –> different people, different places/settings, or different behaviours
(ex. a baseline for how much a student reads at home is set, and then how much that child reads at school is measured)
What are 2 reasons why ABAB Reversal Designs are typically better than ABA Reversal Designs?
- Limits the external error that ABA designs often have
- More ethical to end on treatment (B) rather than withdrawal of treatment (A) as treatment may be beneficial to participant
What are some characteristics of Quasi-Experimental Designs?
- cannot make casual inferences, but indicates how variables are related
- Often used when random assignment is impractical or not possible
- uses preexisting categories/groups
What are the 6 types of Quasi-Experimental Designs?
- One-group posttest only design
- One-group pretest-posttest design
- Non-equivilant control group
- Non-equivilant control group pretest-posttest design
- Interrupted Time Series Design
- Control Series Design
What are multiple repeated measures designs? What designs fall into this category?
Designs that have multiple pretests and posttests overtime rather than just 1
Interrupted time series designs and control series designs are both multiple repeated measures designs
What are the 3 types of developmental designs?
- Cross-sectional
- talk to people at different ages at one time - Longitudinal
- talk to the same people over periods of time
- *Expensive and difficult to do because of how long it takes - Sequential
- 2 groups of people at a cross-sectional time point, and follow both groups over time
- provides a stronger argument for what might be a natural human development!
- a combo of cross-sectional and longitudinal
How do Quasi and Correlational Experiments differ?
Quasi = can deal with multiple (2+) discrete groups without inherent order
Correlational = can deal ONLY with 2 discrete groups without inherent order
**both have no randomization and both measure outcome measures
List and explain the 5 threats to internal validity
- History
- any external event that effects participants between 1st and 2nd measurements
- Ex. midterms are between 1st and 2nd condition of violent video game study - Maturation
- People change over time (become tired, hungry or more mature)
- Ex. mom gives sick son a drink to help him feel better, he feels better a few days later but likely because sickness got better over time not due to the drink - Testing
- Where having a pretest itself can be enough to change a participants protest (like a practice effect)
- Ex. reflecting on how much they smoke in the prestest could be enough to reduce their smoking - Instrument Decay
- basic characteristics of a measurement instrument (timers, software programs, etc.), or the way the participant uses it (become tired, use rating scale differently), changes over time
- Ex. piano recital with rubric example - Regression toward the Mean
- participants who are selected because of their extreme scores tend to later score closer to the mean
- Ex. very hard to get 99% or 1% twice on an exam
What do we try to understand from factorial graphs?
Trying to understand interactions!
What are the steps for gathering info from a line graph?
- Look if the lines are parallel
- If NOT parallel, then there is an interaction! - Look at the midpoints of the lines
- Midpoints = marginal means!
- Marginal means tell you if there is a main effect of the moderator variable (second IV) - Look at the average values between the DV at each level of the IV
- If they are different, it tells you there is a main effect of the primary factor (first IV) that’s on the x-axis
What can we determine depending on if a factorial line graph crosses or not?
IF overlap/crosses –> cannot determine that one IV is always going to lead to higher changes in DV
If NO overlap –> can determine that the IV will always lead to changes in DV
What are simple main effects?
effect of ONE IV on the DV within a single level of the 2nd IV
Ex. More dogs will sit when you tell them to “sit” rather than not telling them to “sit” –
IF you hold food in your hand
^Only one level of second variable = presence of food
^One full effect of first variable = no command, command
Why do we have to decide on one way to split the data for factorial designs?
We cannot analyze data in all possible combinations –> choose the difference that you want to focus on
(usually always the x-axis!)
When choosing a way to split data, what axis is usually always focused on/chosen?
The x-axis (1st variable)
Often trying to see how the moderator (2nd variable) effects that specific chosen interaction on the x-axis!
What are IV x PV designs?
Are a 2x2 factorial design!
allows researchers to investigate how different types of individuals (PV) respond to the same manipulated variable (DV)
Ex.
Factor 1 (iv) = violent vs non-violent video games
Factor 2 (pv) = male vs female
Describe the difference between crossed factorial and nested factorial designs?
crossed factorial = research study that allows for full crossover of ALL possible conditions
nested factorial = one IV is nested in the other IV, so no full crossing
Ex. leaves are nested in trees. Therefore, if a study uses tree species as factor A and leaves as another factor B, factor B (leaves) is nested in factor A (trees)
What set of principles do we use in Canada and what is used in US to evaluate ethical research issues?
Canada = TCPS
US = Belmont Report
List the 3 ethical principles of TCPS (Belmont Report)
- Concern for Welfare (Beneficence)
- Respect for Persons (Autonomy)
- Justice (Justice)
What does Concern For Welfare (Beneficence) involve?
It provides a risk-based analysis
Asks: “Is the participants’ experiences vastly different from their everyday experiences?”
What are the types of risks Concern for Welfare looks at?
- Physical risk (harm to body)
- Emotional/Psychological Harm
- Social Risk (determining if there is a NEED for this study –> would it be ethically wrong for society to not do this study)
- Privacy & Confidentiality
What are exemptions of studies that would not need to meet concern for welfare guidelines?
- program evaluation
- studies that do not involve human/living participants (ex. archival research)
- evaluating teaching method
What does Respect for Persons (Autonomy) involve?
Must treat participants as autonomous people who are able to make deliberate decisions regarding participation.
Why or why not is deception minimal risk?
DECEPTION IS NOT MINIMAL RISK - it’s a violation of respect for persons
What are sins of commission and sins of omission for deception?
Sins of commission (doing something):
- lying
- misleading participants to believe things about experiment or themselves that are not true
Sins of ommission (not doing something):
- leaving out details bc researcher is worried that detail will limit participants desire to participate in study
What must occur with debriefing?
- must explain purpose of research at end of study
- apologize and explain why deception was necessary
- ask participants not to tell anyone!
(debriefing is important with Respect for Persons)
What does the ethical principle of Justice involve?
Fairness and sound rational in participant recruitment
*one population should not bear all the risks of research!
Justice protects disempowered and socially vulnerable populations!
What is assent?
Assent is consent of children given by parents as children cannot express real consent
*This was obviously not received with residential schools –> Indigenous people have a history of experiencing disproportionate costs to research even today
List the benefits of animal research
- allows researchers to control genetic makeup of “participants” –> easier to draw conclusions
- Easier to study physiological, neurological, and genetic foundations of bahaviours.
List and describe the 3 R’s of ethical animal research
- Replacement
- Only do studies on animals if alternatives do not exist - Refinement
- Minimize/eliminate animal distress by modifying experimental procedures
- Includes living conditions - Reduction
- choose animal designs that require the fewest animal subjects as possible
What are some main beliefs animal rights groups hold?
- animals are just as likely as humans to suffer
- animals have inherent rights and values equal to humans
- animals should not bear burden of research to benefit other species
*these all reflect the ethical principle of justice!
Researchers are encouraged to take the perspective of ___________________________
TRUSTING THE PARTICIPANT
What are quantitative variables?
data that can be measured/counted
(NUMERICAL DATA)
What are qualitative variables?
descriptive data of things that cannot be numerically measured/counted
What are scales used for quantitative variables?
- Ordinal Scale
- Ratio Scale
- Interval Scale
Explain ordinal scales
scales where numbers have some sort of meaning and can be placed in an order
Ex. SES, rating an image based on how “ridiculous” it is
Explain interval scales
scales where numbers have a meaning and the difference between each number is maintained throughout the scale
Ex. the amount of energy it takes to go from 31 - 32 degrees Celsius is the same amount of energy it takes to go from 20 - 21 degrees Celsius
Explain ratio scales
scales that are the same as interval scales (numbers have meaning and unit between each number is consistent) but now there is a ZERO
Ex.
- Kevin Scale –> 0 = no kinetic energy
- How much of something your using (possible to be using zero of that thing)
What scales are used for qualitative variables?
- Nominal Scales
Explain nominal scales
scales where the assigned number to a thing is meaningless
*most simple scale there is
Ex. assigning a number to a specific religion group in a study, that number has no meaning to the religion
Define reliability
Reliability is the extent to which we can…
measure something with little error and whether or not a measure is stable over time
Does more error cause more or less reliability?
MORE ERROR = LESS RELIABILITY
LESS ERROR = MORE RELIABILITY
What are 3 ways to measure error?
- Test-Retest Reliability
- Inter-Rater Reliability
- Internal Consistency Reliability
Describe test-retest reliability
assesses similarities between responses at T1 and T2
–> if responses are similar, there is little error
Ex. similar scores of extraversion at T1 and T2 means good test-retest reliability
Describe inter-rater reliability
assesses error based on whether multiple judges/raters responses are consistent with each other
–> if responses between raters are similar, there is little error
How is test-retest reliability evaluated?
Using the Pearson Correlational Coefficient (r)
–> bounded by -1 and 1
How is inter-rater reliability evaluated?
Interclass Correlational Coefficient (ICC)
–> bounded by 0 and 1
Cohen’s Kappa Correlational Coefficient
–> superior to (r) when measuring inter-rater reliability
What are some examples of when inter-rater reliability is used?
- Behavioural Coding (observing interactions)
- Thematic/Content Coding (read content)
- Personality Measures (observe perceived personality traits of targets)
What is internal consistency reliability?
assesses the extent to which ALL items in your measure/questionare assess the same thing
Measured by:
- Inter-item correlation
- Cronbach’s Alpha
- Item-total correlation
What are inter-item correlations?
all separate questions in questionare should be assessing the same overall concept
(looking at coefficients between each item)
Ex. peoples responses to Q4 should be positively correlated to Q5 responses
What is Cronbach’s Alpha?
indicates the extent to which people’s responses to peoples questions are correlated to each other with ONE STATISTICAL NUMBER
*not quite the same as combining inter-item correlations but both are looking at the differences in answers between each question
What is item-total correlation?
the value of overall score should be correlated to the score in each individual question
Ex. if overall score tells you someone is sexist, each response to each Q should reveal sexist tendencies
What is construct validity?
the extent to which we are measuring what we think were measuring
What are types of validity that focus specifically on the items of one measure itself?
- Face Validity
- Content Validity
What are types of validity that focus on comparing measures to other measures and behaviours?
- Predictive Validity
- Concurrent Validity
- Convergent Validity
- Discriminant Validity
What is face validity?
it views each item of a questionare to see if it measures what is intended
*often happens automatically, therefore, not a requirement
What is content validity?
looks at the whole measure to see if it measures what is intended
*questionare should not go beyond what is indended!
Ex. if measuring risk taking, all 4 aspects of risk-taking are being evaluated (thrill-seeking behaviors, rebellious behaviors, reckless behaviors, and antisocial behaviors)
What is predictive validity?
If a test can predict future outcomes
Do people at T1 go on to do relevent behaviours at T2
Ex. If college entrance exam (T1) predict a first semester GPA (T2) that is similar to exam score.
What is concurrent validity?
Sees if people behave in ways you would expect based on their score on a measure in the same time frame
Ex. give person a questionare on environmental conciousness and offer them something to drink, and then see after the questionare if the person recylces their cup or not
What is convergent validity?
if scores on a measure are same to scores on a measure with the same theoretically similar concept
Ex. if scores high on a scale of extraversion, they should score high on a scale of outgoingness to have high convergent validity
What is discriminant validity?
if scores are not related to scores on measures of unrelated constructs (scores should randomly vary)
Ex. if measure reveals that someone scores high on having kind personality traits, scores on measures of colour preferences or car preferences should be randomly varied
What are the main differences between internal validity and construct validity?
internal validity
- focused on how the study is set up
- tries to determine if causality can be inferred
- tries to avoid confounds and alternative explanations
construct validity
- focused on the operationalization of variables
- tries to determine if were measuring what we think were measuring
List reasons for why we use descriptive statistics in general, and for experimental and correlation designs
Generally… to summarize mass of data points and understand/interpret data
In experiments…to calculate central tendency
In correlational designs… central tendency, standard deviation, correlation coefficient
What is THE most important task of descriptive statistics?
To 1) determine what is representative of the data AND 2) determine what is the middle of the data
Describe the mean, how to calculate it, and it’s pros and cons
mean = arithmetic average
- most used measure of central tendency
- uses info from all scores
To calculate: add up all scores and divide by number of scores
Pros:
- maximizes use of data points
- with a bigger sample size, each outlier has less effect on mean
Cons:
- effected by outliers!!
Describe the mode and how it’s calculated
mode = most frequently occurring number
*sometimes no mode, sometimes more than one
To calculate: put all the numbers in order to make a frequency distribution, and then look for the number that occurs the most
**when outliers are present and therefore effect the mean, a mode can often be readily available and used to get a sense of the data distribution
Describe the median and how it’s calculated
median = the number exactly in the middle of the data (where 50% is on one side and 50% is on the other)
To calculate: put all the numbers in order and count to the middle score
*if even data set, take the average of the 2 middle scores!
What is variability?
the spread of distribution of scores (how far apart your data is!)
What are the 3 measures of variability?
- Range (max score - min score)
- Variance (s2)
- Standard Deviation (square root s2)
What does variance help us with?
Variance helps us get the standard deviation –> square root variance to get standard deviation!!
*variance on its own is useless for this class
Explain the rationale for squaring the variance to get standard deviation
If you were to take the average of the values on a distribution to try and get standard deviation, the average would be zero because equivalent positive and negative numbers added together equal 0, and 0 divided by 0 is zero
Sooo… instead, the each positive and negative value that vary on the distribution are squared so there are no more negative numbers. This allows for an accurate average and is then divided by N - 1. Now we have our variance.
In order to get rid of the squaring of values we had to do, the variance is square rooted to get standard deviation.
Define Standard Deviation
- it shows how representative your data is of the mean
- the square root of variance!
What happens to the measures of central tendency in a perfect bell curve?
All measures of central tendency are the same value
(mean, median and mode)
What are the percentages for each standard deviation?
1 SD = 34%
2 SD = 14%
3 SD = 2%
*if scores go beyond 3 SD they are considered outliers
What is correlational coefficient (r)?
a numerical index that shows the degrees of a linear relationship between 2 variables
*just a regular correlation
What is a coefficient of determination (r2)?
the correlational coefficient SQUARED
It measures how well a statistical model (more than 1 IV) predicts an outcome (DV)
A number between 0 - 1, unlike -1 - 1 of (r)
–> the squared takes out negative numbers
If r2 = 0 –> model does not predict outcome at all
If r2 = 1 –> model perfectly predicts outcome (explains everything)
if r2 = between 0 - 1 –> model partially explains outcome
Give an example showcasing coefficient of determination
Marshmallow Experiment
If correlational coefficient between children’s ability to delay gratification at age 5 in the marshmallow test to children’s academic performance at age 15 came to be r = 0.39
r2 = 0.15 –> means that delayed gratification accounts for 15% of academic competence.
What do we use regression for?
to PREDICT behaviour
What do we use descriptive stats and correlations for?
to DESCRIBE behaviour
What are the 2 criteria we must have in order to predict someones outcome on something?
- a relationship between 2 variables
- some “point of origin” (y-intercept)
What is regression?
“correlation on steriods” –> extension of correlation
- much like correlation, regression also measures relationships with variables and does not apply causation!!
- Regression uses the predictor variable (V1) to predict changes in criterion variable (V2)
What are regression models?
- a set of theoretically relevant predictors (IVS) predicting a criterion variable
What is the most important part about regression?
looks at the role of multiple predictors to independently predict the criteria
Uses multiple regression equation! –>
Y = a + b1x1 + b2x2 + b3x3
Compare multiple correlations (R) to Squared multiple correlations (R2)
multiple correlations (R) = type of correlation coefficient that indexes how much a set of predictors (combined) is related to the criterion
^^not useful until… we square it:
squared multiple correlations (R2) = proportion of variability in criterion (DV) account for by a set of predictors (a set of IVs)
What is a partial correlation?
a correlation between X and Y that statistically removes the effect of the 3rd variable
Example:
“Societies where people live the longest have the most cars, therefore cards improve life expectancy”
^^There is a correlation between these variables, but lots of 3rd variables (clean water, health care, etc.)
Partial correlation tries to remove the effects of that third variable so just the correlation of the 2 main variables is presented (cars and life expectancy)
What’s a bivariate correlation?
it means the same as a correlational coefficent (r)
Standardized index of how much two variables change with each other
What is multiple regression?
A technique used when we want to test how well one or more predictors individually predict the criterion
