RM Introduction Flashcards
What order do you plan an experiment
Order:
Theory
Research Questions
Research Design
Hypotheses
Data Collection
What is the experimental design process
Create hypothesis
Translate hyp into treatment conditions or levels
Administer treatment groups to P
Measure performance on response
Treatment conditions
(the variables being manipulated)
are commonly known as…
Independent variables
eg. drug or placebo
amount of caffeine
Response measures are commonly known as…
Dependant variables
eg. blood sugar levels
performance on eye test
Name the 3 types of IV
Quantitative
Qualitative
Classification
Quantitative
Variables represent variation in amount
eg. amount of drug, loudness of noise, difficulty of test
Qualitative
Variables represent variations in kind or type
eg. teaching strategy, type of psychotherapy
Classification
Variables represent characteristics that are intrinsic to the subjects/participants
eg. ex, species, age group
What are Nuisance variables
potential independent variables
if left uncontrolled, could exert a systematic influence on the different treatment conditions
eg. different researchers may produce an “experimenter effect”,
time of day, individual differences
Uncontrolled nuisance variables are also known as…
Confounding variables
– they confound any inference derived from the experiment
What should a good dependent variable should capture
The hypothesised differences
The observed data should be somehow dependent on the independent variable
SCENARIO:
Two labs are practically identical, except that temperature cannot be controlled
Temperature variations may lead to…
Systematic differences in task performance
Solution - random allocation of treatment conditions
(many 0.5, 1.0, 1.5ml caffeine per P) each lab gives an equally likely “chance” that different random temperatures will be associated with the different treatment conditions
A completely randomised design involves…
Random assignment helps to prevent…
Also known as a between subjects or independant-groups design
Each P is randomly assigned to one of the treatment conditions
non-manipulated systematic differences (confounds) from occurring between treatment groups
Observed differences are observed between groups of participants
Randomized block design uses…
Referred to as a repeated measures design or a within-subjects design
Blocks of subjects who are matched closely on some relevant characteristic
A common procedure is to treat a subject as a “block”, wherein the subject serves in all the treatment conditions of an independent variable
A research hypothesis
A fairly general statement about the presumed nature of the world inspiring a specific experiment
eg.
“Physical exercise decreases dementia symptoms”
A statistical hypothesis
A precise statement about the parameters of distributions for different treatment populations
eg. Mean dementia scores will be lower for the Exercise group than for the No-Exercise group
(more than what we expect to observe by chance!)
What 2 types of Statistical hypotheses are stated
Null hyp
Alternative hyp
(mutually exclusive so both should never be true on the treatment parameters)
What does this Greek letter represent
µ
The MEAN
The parameters of the distribution for each treatment population include which 2 things
The mean
µ
The standard deviation σ
Assuming a normal distribution (bell curve),
what represents the length and the width
Length- µ (mean)
Width- σ (standard deviation)
What symbol represents the Null hypothesis
H0
What symbol represents the Alternative hypothesis
H1
A classic statistical test seeks to do what
This is the same as saying that no treatment effects are present in the population
Seeks to accept or reject the null hyp
Generally, what does the equation for the Null hyp express
It expresses the equality between the different treatment populations
What does this Null equation represent:
𝑯𝟎: 𝝁𝟏= 𝝁𝟐= 𝝁𝟑
𝑯𝟎: 𝝁𝟏= 𝝁𝟐= 𝝁𝟑
-for experiment with 3 levels
No treatment effects are present in the population
If the treatment parameters do not agree with the null hypothesis
(i.e., one or more of the differences between treatment means is greater than expected by chance, what must we do
We reject the null hyp in favour of its inverse, the alternative hypothesis (𝐻1 or 𝐻𝐴).
Usually, what does 𝐻1 simply state
the parameters are NOT all equal between treatment populations
When do we decide NOT to reject 𝐻0 (Null) and instead reject 𝐻1 (Alternative)
when our parameter estimates do not differ beyond what would be expected by chance
When we gain a sample from a population
our observed parameters are the…
sample mean 𝒙
sample standard error s
These are approximations to the “true” population parameters µ (mean) and σ (SD)
Why do we use random sampling
To choose a sample that is representative of the underlying population
Population samples may deviate from the population, which is determined by random variation (a.k.a. chance)
To compare existing statistical data with data from our study, what distribution would we use
An F-distribution
This allows us to define chance, by comparing our observed group differences to the theoretical F-distribution
What does the p-value represent
The probability that you could have had a statistical difference due to chance
The lower the p-value…
The less likely 𝐻0 is true
In practice, to decide whether to reject 𝐻0 or not, we quantify differences between groups as an F-value (our test statistic)
What do we do next
(hint: probability p value)
Next, we compute the probability (p-value) of finding this F-value (or greater), given that 𝐻0 is true, by comparing it to the F-distribution
What does the statistical threshold 𝛼 represent
Alpha 𝛼=0.05
If the p-value for a hypothesis test is LESS than 𝛼=0.5, we say that it is what
“statistically significant”
and we reject 𝐻0 (and implicitly accept 𝐻1)
If the p-value for a hypothesis test is MORE than 𝛼=0.5, we say that it is what
NOT statistically significant
and we fail to reject” 𝐻0
(accept the 𝐻0)
The F-test procedure does not guarantee that a correct inference will be drawn
there is a possibility that 𝐻0 is wrong
What are the two basic types of error:
Type I Error (False positive)
Type II Error (False negative)
Type I Error (False positive)
Decision:
We have rejected 𝐻0
and accepted 𝐻1
Reality:
We should have accepted 𝐻0 as
𝐻0 is true (no difference between groups)
𝐻1 is false
Type II Error (False negative)
Decision:
We have failed to reject 𝐻0
and instead rejected 𝐻1
Reality:
We should have rejected 𝐻0 as
𝐻0 is false (real difference between groups)
𝐻1 is true
When you change your alpha threshold…
you are changing the capacity for more chances of a type 1 or type 2 error
What is a true score
A measurement without error
Is approached as the number of items in a test increases.
Increasing length of a test will do what to the reliability
Increase the reliability
since the randomness associated with individual items tends to average out
How do you calculate the
Test score variance?
True score variance
+ The error variance
= Test score variance
eg. True score variance 35 + the error variance 15 = Test score variance 50
How do we calculate the reliability
(proportion of test score variance due to true score variance)
True score variance 35
÷Test score variance 50
= 0.70
Which type of validity is most subjective?
Face validity
Which type of validity is the most theoretically based?
Construct validity
How do you calculate an upper limit of its validity based on its reliability?
The upper limit is the square root of the reliability
Why is experimental control used?
To handle nuisance variables
that confound scientific inference
eg. randomised design
Psychologists’ goal is to minimise the chance that the decision is what
erroneous (type 1 or type 2)
A researcher does a survey randomly calling phones that have land lines.
People who only have cell phones are not sampled.
This is an example of…
Undercoverage bias
since those with only cell phones are not only undercovered but not covered at all
A radio station asks readers to phone in their choice in a daily poll. This is an example of…
Self-selection bias
since those with strong feelings are most likely to respond
A researcher surveys people who have been in therapy for 5 years with the same psychotherapist. This is an example of…
Survivorship bias
Those who stay for 5 years may be more satisfied with their therapist than average.
May also have more severe problems if they stay in therapy so long
If treatment is successful more in women than men
What is the interaction?
The interaction is between gender and treatment
The sum of the effects of the unmeasured variables can be estimated by…
by the within-group variances
Differences within a group are due to unmeasured variables
so the variances within the groups can be used to estimate the effects of the unmeasured variables
