test 1 key things Flashcards
Core Characteristics of Science:
- Systematic Empiricism:
Systematic: rigourous application of the
scientific method to collect valid and
reliable data about our observations of the
world and allows research to be open for
criticism and peer review. A standardised
scientific methodology ro reduce bias.
Empiricism: learning through the senses
(experiences; including scientific tools)
rather than from logic or authority. - Addresses Empirical Questions:
Scientific research engages in hypothesis
testing; falisifiable, on topics we can
collect data for, is ethical, not too broad or
narrow and provides novel insights. - Creates Public Knowledge:
Findings are available to, communicated
and open to critique by the public.
Review and replicatio of studies findings
by other scientists to rule out alternative
explanations, replicate findings on other
samples and extend theory.
*biggest difference between science and
authority is its acceptance of critiscims and
challanging perspective.
*it’s not “what” you study but “how” you
study.
*public trust in science has increased since
the introduction of mathematical modelling
and developing of vaccines.
How Do We Know What We Know?
Three types of knowledge:
- Experience:
Information obtained from our own life
experiences.
Its unfalsifiable.
Its subjective; not everyone has the same
experiences, can have the same
experiences or wants to have the
experience.
Prone to memory biases (i.e., reconstruction
of events).
Limited sampling. - Intuition:
Unspoken understanding and assumptions
about how the world works.
Its implicit; so occurs outside of our
unconscious awareness so is not guided by
our existing knowledge or open to criticism;
prone to biases; e.g. myths.
Its unfalsifiable.
Its subjective. - Authority:
Information obtined from an authority
figure/source; teacher, textbook, parents,
school, political leaders, spiritual leaders
etc.
Is often accepted without consideration for
its validity.
The truth can be weilded to the authority
figures will and biases they may hold.
Lacks falsifiability.
May be biased.
theory, hypothesis, predicition
Theory: o A possible explanation of a broad range of phenomena. Hypothesis: o A more specific explanation of how something might work. o About the construct of interest. Prediction: o What will happen in my study if my hypothesis is right. o A specific prediction on the relationship between two or more variables. o Operationionalized variables which indirectly measure the construct of interest.
*the cycle of deductive reasoning used in
experimental designs which builds on previous knowledge.
Two Ways of Knowing:
*both valid approaches
Inductive Reasoning:
o Building a theory/hypothesis from patterns
we see in observational data.
Deductive Reasoning:
o Testing our theory/hypothesis by finding
patterns in observational data.
Three Types of Scientific Claims:
- Describe:
- Frequency claim typically about one variable
at a time.
- Descriptions about how many, how much,
how often and what kinds can be either
quantitiative or qualitative:
o Qualitative:
Coding, thematic anlaysis, enthnography or
diaries are common techniques used in
naturalistic observational work.
o Quantitative:
Mean, median, mode, frequency, standard
deviation, percentiles, z-scores are
quantitiative tools used to describe the
shape of the data. - Predict:
- Association claim about two or more
variables.
- Good way to validate your hypothesis that
there is an association between the
variables.
- Variables are measured, they are not
manipulated.
- Individual group differences with no
manipulation, associations or correlations
can be used.
- Correlation/Association ≠ causality. - Explain:
- Provides causal explanations of why or how
things occur.
- Manipulation of the independent varaible (IV)
to measure its effects on the dependent
variable (DV).
- This allows us to determin whether the (IV)
causes changes in the DV, by controlling all
other variables.
- Claims about the direction of the relationship
between variables.
examples: § How would we make this claim into a hypothesis/prediction? o Frequency How many people are practicing mindfulness? How many people are anxious? o Association Are people who practice mindfulness less anxious? Are people who are mindful in their daily lives less anxious? o Causation o Does the practice of mindfulness reduce anxiety? o How does mindfulness reduce anxiety?
Use The Right Language:
Association:
- Correlated with
- Associated with
- Related to
- Relationships between
- X predicts Y (predicts ≠ cause)
- Group differences without a manipulation
Causal:
- Causes/makes/affects
- Increases/decreases
- Heightens/reduces
- Attenuates/potentiates
- Eliminates (bold claim)
- directional
- action verbs
three conditions for causality
Three Conditions for inferring causation:
1. An association between mindfulness and
anxiety (correlation; bidirectional)
2. Temporal precedence (cause before
effect) IV causes changes in DV.
3. Control of alternative explanations
(confounds) keep extraneous constant and
manipulate the IV!
Experiments:
All experiments have 2 components (defining features).
Experiments:
All experiments have 2 components (defining features).
- Manipulation
- of one or more independent variables
(cause) to determine its effect on a
dependent variable. - Control
- of alternative explanations (extraneous
variables)
*This allows us to fulfill the three conditions from inferring causation!
Variables
Variables:
1. Independent variables (IV)
- Manipulated by the experimenter
- The Cause.
2. Dependent variable (DV)
- Measured by the experimenter
- The Effect.
3. Subject variables
- Pre-existing factors (age, gender,
ethnicity, personality, etc)
- NOT MANIPULATED
4. Extraneous variables
- All the other variables that contribute to
the DV
- Example for things that contribute to
anxiety and will always be present in a
study: gender, SES, genetics,
environment, social media, physical
health, temperament, trauma/stressful
life events etc.
- Random assignment allows for these
factors to be equal across conditions-
the only difference between groups
should be the level of the independent
variable they experience.
*if not manipulated we cannot infer cause!
§ Subject Variables:
- Might be interested in whether different
types of participants are affected by your IV
differently.
- BUT – you can’t manipulate these variables,
so they are not independent variables. And
you cannot make causal claims about them.
o Demographic variables (gender, age,
culture or ethnicity, income…)
o Traits (personality, abilities,
o Disorders (depression, anxiety, eating
disorders…)
o Life experiences (travel, major, lifestyle
choices…)
§ Extraneous Variables: - All other possible variables that could contribute to variability in the DV. § Confounds: - Extraneous variables that vary systematically with the IV. - Confounds create alternative explanations for findings.
Recap
§ We get rid of confounds by designing better
conditions. This allows us to infer causation
and have stronger internal validity within our
study.
§ We can satisfy the three conditions of
causality through the use of manipulation
and control. Manipulation helps us to
determine the direction of relationship
(temporal) and control allows us to rule out
alternative explanations or confounds.
§ All studies have limitations and are always
reported at the end of a scientific article.
§ No study is perfect. There are tradeoffs
between all decision researchers make
when designing and implementing an
experiment that will place limitations on the
study.
§ Just because a study is flawed doesn’t mean
that its findings are useless.
§ We cannot answer all questions in one
study. There are multiple ways to
operationalize IV-DV’s and other
methodological choices to answer the same
question which all provide a useful piece to
the puzzle.
§ You cannot make a conclusion from one
study; we need different studies that adopt
different perspectives, methodologies and
flaws which all draw the same conclusion to
draw a conclusion and be confident the
effect is present (converging
operations/consensus). More important to
be aware of your studies flaws than to have
none. Some studies methodologies can be
so flawed that their findings are not valuable
in the scientific community- lab example or
non-scientific studies run by businesses.
Psychology is valuable in any area or
industry of life; critical analysis of findings
and the implementation of non-scientific
studies (i.e., education; ministry of health).
Random Assignment:
§ All participants have an equal likelihood of
being assigned to each experimental
condition.
§ If groups are large enough, controls for
differences in all extraneous variables
(equally distributed across conditions)
§ Assignment must be truly random. Failures
of random assignment occur when:
o People choose their own conditions
o People are assigned to condition based on
some preference/trait/behaviour
o There are systematic differences between
groups.
§ Any differences between groups other than
the IV are confounds.
§ What is the difference between an
extraneous variable and confound?
o An extraneous variable is all other possible
variables that could contribute to variability
in the DV.
o Confund variables are extraneous variables
which differs on average across the levels of
the independent variable[s] (i.e.,
intellegence if there is not an equal mix of
high and low IQ participants in each
condition).
§ Example of people are assigned to condition
based on some preference/trait/behaviour:
o Are executive monkeys more likely to
develop stress induced ulcers relative to
employee monkeys?
o Monkeys were trained on pressing a lever to
stock minor electrical shocks from occurring
in their cage. They took the first 6 monkeys
to learn the association and made them the
executives. The rest were the employees
that were at the mercy of the executive to
press the lever at the correct time to stop
the shock. They concluded that executive
monkeys with the additional responsibility
experienced more stress and developed
ulcers relative to employee monkeys.
o We now know that powerlessness is the
bigger cause for ulcers than
power/responsibility is.
o This is not random assignment because
monkeys were assigned to a group based
on a subject variable. There was no
manipulation! Or equal probability of being
assigned to a condition.
Confounds:
§ Extraneous variables that vary
systematically with the IV.
§ Confounds create alternative explanations
for findings.
§ For example, differences between the
mindfulness and waitlist-control condition:
o Expectations of improvement.
o Time commitment.
o Relationship with group leader.
o Relationship with other group members.
o Relaxation.
§ What would be a better control group to
use than a waitlist control group:
o Mindfulness training and mindfulness
training without all those other things.
o Mindfulness treatment vs alternative
treatment.
§ Different control groups will address
different confounds. Thus, as a researcher
we need to decide which one’s are most
important for us to control to ensure the
study has high internally validity and we
can be confident that changes in the DV
are caused by changes in the IV.
§ Gina’s example:
o Use an “active control-group” i.e., an
experiment where there are two treatment
conditions.
o For example,
§ Mindfulness vs home improvement class
(with structure and social aspects but
doesn’t include mindfulness training).
§ Mindfulness vs CBT (alternative treatment).
Factorials why add another variable?
Why add another variable?
1. Save time. Assess 2 or more potential
causes at once (relative to multiple studies)
2. To refine a theory (because it depends… to
identify what situational factors the effect
of the IV depends on).
3. To rule out confounds.
4. To increase external validity (extend to
other populations, stimuli, situations…;
good for quasi-experimental designs).
Summary
Summary:
Experimental Designs 1. Post-test only 2. Pre-test/Post-test 3. Matched Pairs 4. Within-Subjects 5. Factorial Designs Independent variables • Manipulated • One or more IVs • Each with 2 or more levels • Manipulated between- or within-subjects • Usually categorical, but might reflect an underlying continuous variable
Dependent variables
• Measured
• Can be categorical or continuous
• Can include more than 1 DV in a study
calculate t
1) Calculate the test statistic (t for t tests)
• Test statistic
• group difference divided by standard error
• = ratio (variability between groups
explained by IV/natural variability in
sample-sampling error-difference from
population to mean-confidence it’s close to
true mean)
• 28/128
• t = 0.218
Q: How often would I get a t of this big or bigger if the null hypothesis is true?
• Is answered comparing the t= 0.218 with
the sampling distribution.
2) Compare it to the sample distribution
• The sampling distribution tells us how likely
it is to get the same t-score if the null
hypothesis is true (or bigger when
sampling randomly from the same sample).
• Sampling distribution shows me how often I
will get different values of a test statistic
(e.g., t) by randomly sampling from a single
population.
• Shape of the sampling distribution depends
on how big my sample is (for an
independent t-test, degrees of freedom = N
– 2)
• The larger the N the more likely given the
null hypothesis is true should the
differences between groups should be
smaller i.e., closer to 0.
• In the tail of the sampling distribution, the
rejection region where it is not likely that
the test statistic was produced by the null
hypothesis being true (sampling error and
not due to IV)
• T-score that is closer to 0 means it’s more
likely to be caused by null and not IV,
Bigger Test scores are better! In tails.
• Shape of the t distribution depends on the
number of people in your sample. The more
people, the more likely t will be close to 0
(i.e., more like a normal distribution).
• The shape changes based on DF, which is
in turn, impacted by N size and group # (i.e.,
20 – 2 = 18).
- Reject null: if the likelihood of the null
hypothesis being true is less than 5%
(produce test statistic this size or greater
less than 5% of the time). - Bidirectional (5% reject, two tail, 2.5% each
side; need bigger t to fit in small tail or more
evidence to reject)
-Directional (5% reject, one tail 5% on one
side, smaller t is needed, less evidence
needed, divide p-value by two to account
for bigger rejection region)
