quantitative research Flashcards
psychology
the scientific study of behaviour and mental processes
criteria for a scientific study
● It should be supported by empirical evidence
and be based on this evidence.
● It should be falsiable, that is, it should be
possible for the theory or study to be proven
wrong.
● There should be a history of independent
attempts to test the theory or replicate the
study.
artifect
It was recognized that
experiments, if not carefully controlled, could
produce artifacts—results that are associated with
the effect of unforeseen factors.
behaviour
Behaviour is everything that can be registered
by an independent observer: it includes overt
actions as well as gestures, facial expressions,
verbal responses, endocrine reactions and so on.
mental processes
such as attention, perception, memory
and thinking. We cannot observe them directly but we can observe the indirect
effects mental processes have on one’s behaviour.
So, we can infer something about the mental
world as well.
in what form does data in quantitative come in
Data in quantitative research comes in the form
of numbers. The aim of quantitative research is
usually to arrive at numerically expressed laws
that characterize behaviour of large groups of
individuals (that is, universal laws).
In philosophy of science such
orientation on deriving universal laws is called the nomothetic approach.
how does quantitative research operate
Quantitative research operates with variables.
A variable (“something that can take on varying
values”) is any characteristic that is objectively
registered and quantied.
operationalisation
Operationalization of a construct
means expressing it in terms of observable
behaviour.
construct
A construct is any theoretically dened variable,
for example, violence, aggression, attraction,
memory, attention, love, anxiety. To dene
a construct, you give it a denition which
delineates it from other similar (and dissimilar)
constructs. Such denitions are based on
theories. As a rule constructs cannot be directly
observed: they are called constructs for a
reason—we have “constructed” them based on
theory.
3 types of quantitative reaseacrh
- experimental studies
-correlational studies
-descriptive studies
experimental study
-The experiment in its simplest form includes one independent variable (IV) and one dependent variable (DV), while the other potentially important variables are controlled.
-The IV is the one manipulated by the researcher.
-The DV is expected to change as the IV changes.
-only method that allows cause and effect inferences
correlational studies
-Correlational studies are different from experiments in that the researcher does not manipulate any variables (there are no IVs or DVs).
-Variables are measured and the relationship between them is quantied.
-However, you cannot make cause-and-
effect inferences from correlational studies.
-Since you did not manipulate one of the
variables, you do not know the direction of
inuence.
descriptive study
In descriptive studies
relationships between variables are not
investigated, and the variables are approached
separately.
aim, data, focus and objectivity of quantittaive research
aim: nomothetic apprahc, derive universally applicable laws
data: numbers
focus: behavioural manifestations (operationalisations)
objectivity: more objective as the researcher is eliminated from the studied reality
aim, data, focus and objectivity of qualitative research
aim: Idiographic approach: in-depth
understanding of a particular case or
phenomenon
data: texts
focus:Human experiences, interpretations,
meanings
objectivity: more suvjective
a sample and sampling
A sample is the group of individuals taking part
in the research study. Sampling is the process of
nding and recruiting individuals for the study.
credibility
Credibility refers to the degree to which the results of the study can be trusted to reect the reality. It is closely linked to bias, because the results of the study do not refect reality if there was some sort of bias in it.
generalisability
Generalizability refers to the extent to which
the results of the study can be applied beyond
the sample and the settings used in the study
itself.
sampling types in experimental studies
Random
Stratied
Self-selected
Opportunity/convenience
sampling types in correlational studies
Random
Stratied
Self-selected
Opportunity
sampling types in qualitative research (5)
Quota sampling
Purposive sampling
Theoretical sampling
Snowball sampling
Convenience sampling
generalisability in experimental studies
External validity:
– Population validity
– Ecological validity
Construct validity
generalisability in correlational studies
Population validity
Construct validity
generalisability in qualitative research
Sample-to-population
generalization
Case-to-case generalization
Theoretical generalization
credibility in experimenyal studies (table)
Internal validity: to what extent is the DV inuenced by the IV and not some other variable?
Controlling confounding variables: eliminating or keeping constant in all conditions
credibility in qualitative research
Credibility = trustworthiness. To what extent do the ndings reect the reality?
-Triangulation
-Establishing a rapport
-Iterative questioning
-Reexivity
-Credibility checks
-Thick descriptions
credibility in correlational studies
(table)
No special term used: “validity” and “credibility” can be used interchangeably
Credibility is high if no
bias occurred
bias in ex studies
Threats to internal validity:
– Selection
– History
– Maturation
– Testing effect
– Instrumentation
– Regression to the mean
– Experimental mortality
– Experimenter bias
– Demand characteristics
bias in corr. studies
On the level of measurement of
variables: depends on the method of measurement
On the level of interpretation of
ndings:
– Curvilinear
relationships
– The third variable
problem
– Spurious
correlations
bias in qualitative research
Participant bias:
– Acquiescence
– Social desirability
– Dominant respondent
– Sensitivity
Researcher bias:
– Conrmation bias
– Leading questions bias
– Question order bias
– Sampling bias
– Biased reporting
confounding variable
Variables that can potentially distort the relationship between the IV and the DV are called confounding variables.
They contribute to bias. These variables need to be controlled, either by eliminating them or keeping them constant in all groups of participants so that they do not affect the comparison.
target population
The target population is the group
of people to which the ndings of the study are
expected to be generalized. The sample is the
group of people taking part in the experiment
itself.
random sampling
- This is the ideal approach to make the sample representative.
- In random sampling every member of the target population has an equal chance of becoming part of the sample.
- With a sufcient sample size this means that you take into account all possible essential characteristics of the target population, even the ones you never suspected to play a role.
- Arguably, a random sample of sufcient size is a good representation of a population, making the results easily generalizable.
- However, random sampling is not always possible for practical reasons.
representativeness
A sample is said to be representative of
the target population if it reects all its essential
characteristics.
stratified sampling
- theory-driven. First you decide the essential characteristics the sample has to reect.
- Then you study the distribution of these characteristics in the target population
- Then you recruit your
participants in a way that keeps the same proportions in the sample as is observed in the population randomly or use other approaches - it ensures that theory-dened essential characteristics of the population are fairly and equally represented in the sample.
- This may be the ideal choice when you are certain about essential participant characteristics and when available sample sizes are not large.
convenience sampling
- For this technique you recruit participants that are more easily available.
- There could be several reasons for choosing convenience sampling.
- First, it is the technique of choice when nancial resources and time are limited.
- Second, there could be reasons to believe that people are not that different in terms of the phenomenon under study.
- Finally, convenience sampling is useful when wide generalization of ndings is not the primary goal of your research,
self-selected sampling
- This refers to recruiting volunteers.
- The strength of self- selected sampling is that it is a quick and relatively easy way to recruit individuals while at the same time having wide coverage
- The most essential limitation, again, is representativeness.
- People who volunteer to take part in experiments may be more motivated than the general population, or they may be looking for the incentives
independent measures design
-involves random allocation of participants into groups and a comparison between these groups.
- In its simplest form, you randomly allocate participants from your sample into the experimental group and the control group.
- Then you manipulate the experimental conditions so that they are the same in the two groups except for the independent variable.
- After the manipulation you compare the dependent variable in the two groups.
- The rationale behind random group allocation is that all potential confounding variables cancel each other out.
- If the groups are not equivalent at the start of the experiment, you will be comparing apples to oranges.
- Conversely, when the group sizes are sufciently large and allocation is completely random, chances are that groups will be equivalent—the larger the sample, the higher the chance.
MAtched pairs design
similar to independent
measures. The only difference is that instead of
completely random allocation, researchers use
matching to form the groups.
The variable that is controlled is called the matching
variable.
Matched pairs designs are preferred when:
● the researcher nds it particularly important
that the groups are equivalent in a specic
variable
● the sample size is not large, therefore there is a
chance that random allocation into groups will
not be sufcient to ensure group equivalence.
repeated measures design
- Repeated measures design is used when the goal is to compare conditions rather than groups of participants.
- The same group of participants is exposed to two (or more) conditions, and the
conditions are compared. - The problem with repeated measures designs is that they are vulnerable to order effects:
- resultsmay be different depending on which condition
comes rst - Order effects may
appear due to various reasons, such as the following.
● Practise: participants practise, improve their on-task concentration and become more comfortable with the experimental task during
the rst trial. Their performance in the second
trial increases.
● Fatigue: participants get tired during the rst trial, and their concentration decreases. Their
performance in the second trial decreases. - An advantage of repeated measures designs is that
people are essentially compared to themselves,
which overcomes the inuence of participant variability - Another advantage following from this is that smaller sample sizes are required.
how to overcome order effects
To overcome order effects researchers use
counterbalancing. Counterbalancing involves
using other groups of participants where the order
of the conditions is reversed.
construct validity
Construct validity characterizes the quality
of operationalizations.Moving from an operationalization
to a construct is always a bit of a leap. Construct
validity of an experiment is high if this leap is
justied and if the operationalization provides
sufcient coverage of the construct.
internal valifity
Internal validity characterizes the methodological
quality of the experiment. Internal validity is high
when confounding variables have been controlled
and we are quite certain that it was the change in
the IV (not something else) that caused the change
in the DV. In other words, internal validity links
directly to bias: the less bias, the higher the internal
validity of the experiment.
population validity
Population validity refers to the extent to which ndings can be generalized from the
sample to the target population. Population validity
is high when the sample is representative of the
target population and an appropriate sampling
technique is used.
ecological validity
Ecological validity refers to
the extent to which ndings can be generalized
from the experiment to other settings or situations.
It links to the articiality of experimental
conditions. In highly controlled laboratory
experiments subjects often nd themselves in
situations that do not resemble their daily life.
external valifity
External validity characterizes generalizability of
ndings in the experiment. There are two types of
external validity: population validity and ecological
validity.
selection bias
-Occurs if the groups are not equivalently controlled for variable besides the independent variable thats supposed to change at the start of the experiment
-Therefore, inferences cannot be made about the influence of the IV on the DV
-Usually happens because case group allocations were not truly randomised
history bias
-The outside events participants experience during the course of the experiment
-May affect the internal validity as the dependent variable may change as a result of the event
-The presence of confounding variables may be controlled, in an experiment by either controlling or eliminating them
maturation bias
-the natural developmental processes participants go through over the course of the experiment
-May affect internal validity as there is no telling whether the IV affected the final results of natural growth and development
-May be counteracted by using a control group (a group that does not experience the enforcement of the IV and is the same in every other respect)
testing effect
-The first measurement of the dependent variable may affect the second and so on and so forth
-Thus, it is important to have a control group to discern what results are from the testing effect and what results are from the changing the IV
instrumentation bias
-Occurs when the instrument measuring the DV results changes slightly between measurements
-Especially poignant in psychology as most instruemnts of measurement are human observers
-Can be avoided by setting standardized measurement conditions accross all comparison groups and observers
mortality
-Refers to the fact that some participants drop out of the experimeent before it has concluded, which may cause problems if the dropouts are not random, which would lead to a lack of equivalence between controlled and experimental groups
-There is no advised way to combat mortality, except creating an environment that does not prompt dropping out
regression to the mean
-more intriguing source of bias that increasingly prevalent in experiments where initial DV score is an extremely low or high
-As extremes have a statistical tendency to average out as the trials continue or as trial repeat. This is true for any trial in which luck in involved
-A countermeasure one may take is a control group with the same baseline starting point and measurements of the same point as well, however, without intervention
-it it very common to infer causation when regression to the mean is actually present
demand characteristics
-This refers to when particpants become aware of the aim of the experiment and (subconciously) chang etheir behabiour according to fit their interpretation
-A viable countermeasure in many cases could be deception, in which you make it difficult for particpants to discern the aim of the experiment (but this may raise ethical concerns)
-Another countermeasure may be a post-exxperimental questionnaire to find out the degree of involvement of deman characteristics in the results
-The threat of demand characteristics increases in repeated measures designed experiments
experimenter bias
-Occurs in a situation in which the researcher unintentionally exerts an influence over the results of the study
-A countermeasure for experimenter bias is double blind designs in which information which could induce subconcious bias is withhels from not only participants but also researchers
quasi experiments
- Quasi-experiments are different from “true” experiments in that the allocation into groups is not done randomly.
- Instead some pre-existing inter-group difference is used. - - The major limitation of a quasi-experimental design is that cause-and-effect inferences cannot be made.
- This is because we cannot be sure of the equivalence of comparison groups at the start of the study
- pre-existing differences in one variable may be accompanied by
- In the way they are designed (supercially) quasi experiments resemble “true” experiments, but in terms of the possible inferences (essentially) they are more like correlational studies.
field experiment
- Field experiments are conducted in a real- life setting.
- The researcher manipulates the IV, but since participants are in their natural setting many extraneous variables cannot be controlled.
- The strength of eld experiments is higher ecological validity as compared to experiments in a laboratory.
- The limitation is less control over potentially confounding variables so there is lower internal validity.
natural experiment
- Natural experiments, just like eld experiments, are conducted in participants’ natural environment, but here the researcher has no control over the IV—the IV occurred naturally.
- Ecological validity in natural experiments is an advantage and internal validity is a disadvantage owing to there being less control over confounding variables.
- Another advantage of natural experiments is that they can
be used when it is unethical to manipulate the IV
-all natural experiments are quasi experiments as the researchers do not manipulate any IV’s
what is correlation
A
correlation is a measure of linear relationship
between two variables. Graphically a correlation is
a straight line that best approximates this “cloud”
in the scatter plot.
pos and neg correlations
A positive correlation demonstrates the tendency
for one variable to increase as the other variable
increases. A negative correlation demonstrates
the inverse tendency: when one variable
increases the other variable decreases. The
steeper the line, the stronger the relationship.
effect size
The absolute value of the correlation coefcient
(the number from −1 to 1) is called the effect size.
statistical significance
- Statistical signicance shows the likelihood that a correlation of this size has been obtained by chance.
- It depends on the sample size:
with small samples you cannot be sure that an obtained correlation, even if it is relatively large, has not been obtained due to random chance. - With large samples correlation estimates are more reliable and you can be more condent that the correlation is not a product of random chance but a genuine reection of a relationship between the two variables in the population.
- The probability that a correlation has been obtained due to random chance can be estimated.
lmitations of correlational studies
- correlations cannot be interpreted in terms of causation
-the third variable problem:There is always
a possibility that a third variable exists that
correlates both with X and Y and explains the
correlation between them.
-curvilinear relationships:
-spurious correlations
spurious correlations
- When a research study involves calculating multiple correlations between multiple variables, there is a possibility that some of the statistically signicant correlations would be the result of random chance.
- When we calculate 100 correlations and only pick the ones that turned out to be signicant, this increases the chance that we have picked spurious correlations.
curvilinear relationships
- Sometimes variables are linked non-linearly.
- However, this relationship can only be captured by looking at the graph.
- Since correlation coefcients are linear, the best they could do is to nd a straight line that ts best to the scatter plot.
-Psychological reality is complex and there are a lot of potentially curvilinear relationships between variables, but correlational methods reduce these relationships to linear, easily quantiable patterns.
credibility in correlational research
Bias in correlational research can occur on the
level of variable measurement and on the level of
interpretation of ndings.
On the level of measurement of variables, various
biases may occur and they are not specic to
correlational research.
