CH 4 & 8 - Research and experimental design Flashcards
AFter formulating the research question you should…
Formulate the hypothesis
Ater formulating a hypothesis you should…
define your variables
After defining your variables you should…
choose your measurements
After choosing your measurements you should…
design the study
After designing the study you should…
Collect the data
After collecting the data you should…
Analyze the data
Research questions should be
Clear, focused, concise, complex, arguable
Clear research questions…
provides enough specifics that one’s audience can easily understand its purpose
without needing additional explanatio
A focused research question is
narrow enough that it can be answered thoroughly in the space the writing task allows
A concise research question is
expressed in the fewest possible words
A complex research question is
not answerable with a simple “yes” or “no,” but rather requires synthesis
and analysis of ideas and sources prior to composition of an answer
An arguable research question means…
its potential answers are open to debate rather than accepted facts
Directional hypotheses
Specifies what we expect the relationship between two variables to be
Non-directional hypotheses
Specifies the existence of the relationship without making assumptions about the nature of this relationship
The opposite of an alternative hypothesis is…
A null hypothesis (no relationship between variables)
Does rejecting the null hypothesis prove the alternative hypothesis?
No. It only supports it
What does it mean to support the alternative hypothesis, in statistical terms?
Support means that we found out there is a very small chance to find the results we have assuming the null hypothesis is true
What are the three general categories of variables?
Situational, Response, Participant/classificatory
Situational Variable
Characteristic/event a participant is exposed to - characteristic of situation/ environment (e.g. IV)
Response variable
Participant’s reaction/behavior (e.g. DV)
Participant/classificatory variable
Pre-existing aspect of participant that is of interest
0 = r means that
There is no correlation
Non-experimental reseach
lacks manipulation of ID and results in descriptive or correlational results
Criteria for claiming causality
Covariation, temporal precedence, internal validity
Covariance
If x changes, so does Y
Temporal precedence
the cause comes before effect
Directionality problem
Directionality problem
Before an experiment We don’t know which variable is the cause and which one is the effect
Internal validity
How much an evidence supports a claim. Must RULE OUT alternative explanations to be high
Categorical independent variable
IV has several levels that correspond with experimental conditions (also called treatments) - e.g. medication 1, 2, control
Continuous independent variable
IV is a spectrum. (e.g. height)
Categorical dependent variable
DV is divided into categories and person’s measurements falls into one of them. (e.g. if person passed a test or not)
Continuous dependent variable
DV has no categories and is a spectrum. Person’s measurements falls on a scale (e.g. seconds to perform a task)
Types of dependent variables
Self report
Physiological (biological data)
Behavioral (observed by scientist)
Third variable
Variable that causes/impacts BOTH variables of interest (E.g. T causes X and Y)
Confounding variable
Variable that CO-VARIES with a variable of interest
Based on operational definition, IMPOSSIBLE to separate from the variable it covaries with
Can explain part of the results
(e.g X + C cause Y)
Mediator variable
A process explains the relationship between two variables (e.g. X causes M that causes Y)
How to test for mediation
- Test if X causes Y
- Test if X causes M
- Test if M causes Y
- Run a regression test to see if X or M predict Y better. See if X causing Y disappears.
- Establish temporal precedence
Random assignment
Randomly put participants into the different levels/conditions
No systematic reasons for why participants are in certain conditions
More effective as number of participants increase
Random variables are balanced out
Systematic variability
When levels of a variable coincide (in a predictable way) with group assigned membership- creates potential confound
EX: (like if situation 1 has worse equipment than situation 2)
Unsystematic variability
Individual differences withing group. Random. Variable fluctuates independently of group assigned.
Control vs experimental groups must be
Identical, except for the IV
Between subject design
Participant is tested in one condition ONLY
Within subjects design
Each participant is tested under all conditions
Posttest only design
IV groups are set and later DV is measured
Pretest/Posttest design
Measure DV before test,
IV groups are divided (into different conditions)
Complete the same DV measurement again
Posttest only Pros and Cons
There is no pretest to sensitize participants to the hypothesis
but Can’t measure change over time
Pretest/Posttest
Useful when interested in changes over time and when you want to be certain the groups are equivalent at the start (baseline)
BUT must be careful to not make the participants change their spontaneous behavior
Matched groups/ pairs (or yoked design
Similar participants (in relation to variable) are grouped into sets. Each member is matched with a similar person, and the pairs are randomly split into the conditions.
Repeated measures design
Within-subjects design
Participant is exposed to 1 IV condition, is tested (DV), is exposed to 2nd IV condition, Is tested again (DV)
Concurrent-measures design
Within subjects
Participant is exposed to all levels of IV at the same time. Dependent variable is measured based on the behavior or attitude response from the double exposure.
Couterbalancing
Repeated measures design –> order can affect results. Counterbalancing controls order effects by presenting the levels of IV in different sequences (randomly assign participants to each sequence)
Full counterbalancing
All possible condition orders are represented - there is a IV level for each possible order
Partial counterbalancing
There is an IV level for SOME possible exposure sequences (e.g. latin square: one level is first in at least one condition)
Within subject designs PROS
No group differences (because there is no control group and experimental groups)
Each participant is their own control
More statistical power
Requires few participants
Power (in an experiment)
The likelihood that a study will show statistically significant results
Order effects
Threat to internal validity in a within-subjects design, in which exposure to one condition changes participant responses to a later condition
Order effects - Practice effects
Participant performance improves over time due to practice of the measurement
Order effects - Fatigue effects
Participant’s performance gets worse because they get tired
Order effects - Carryover/Contrast effects
When contamination carries over from one condition to the next (e.g. understanding what is being tested)
Within subjects designs CONS
Creates a potential for order effects
Unpractical
Might be impossible
More likely to create demand characteristics (knowing what the study is about and wanting to succeed)