RESEARCH
METHODS – NON EXPERIMENTAL METHODS:
Self report
techniques (Questionnaires and Interviews)
+ They hold
the advantage that you can access what people think. You can also use
observation; you can make guesses from how the person behaves.
-
It
may be unreliable as the person might choose not to give you a valid answer
Good
questions in a Questionnaire:
1.
Clarity questions need to be written so that
the reader understands what is being asked of them. One way to do this is to operationalise
certain terms = make sure there is no ambiguity
2. Bias any bias questions may lead to the
respondent giving a certain answer. Need to avoid any leading questions. Main
problem is social desirability bias – people prefer to give answers that make
them look like a good person may not be truthful
3.
Analysis questions should be designed with
analysis in mind. Once a researcher has collected the data the answers need to
be summarised s that conclusions can be made.
-
Filter questions (include some irrelevant questions to mislead ppt)
-
Sequence of questions(start with easy questions do hard at end so they don’t get
anxious)
-
Pilot (allows you to test questions and ensure there is no bias, ambiguity
and that it is easy to analyse)
Quantative and Qualitive
analysis
Qualitive data = WORDS
Quantitive = NUMBERS
For quantitive data you need closed questions which
have a limited number of answers
+it
is easier to analyse
-
Respondents may be forced to select answers that do not
represent their real thoughts
For qualitive data (data that expresses the quality of
things) you need open questions
+can
provide rich, unexpected data on a topic where researchers can learn new things
-
It’s more difficult to analyse than quantitive data. Harder
to detect patterns
Interviews
Interviews are a questionnaire held
face to face or on a telephone. It can be structured or unstructured. In a
structured interview the questions are decided in advance. In a semi-structured
or unstructured interview some or all of the questions are developed at the
interview.
+questionnaires
can be difficult for some e.g. children who find writing difficult. Interviews
are a better data collection method for them
+an
experienced interviewer can get lots of information through the use of
questioning
+semi
structured or unstructured interview, more data can be gathered as questions
can be changed to fit the respondent
+interviewer
can clarify more ambiguous questions or give more information if needed
-
Person may choose to give an invalid answer people may feel uncomftable
revealing personal information face to face, may prefer to write to down
-
Interviewer bias certain words may be empised and this could give a biased
answer
Issues of validity and
reliability in self report techniques:
§
Social desirability bias
§
Interviewer bias
§
Leading questions
§
Content validity
Assessing validity
§
Lie scale (one way to asses if participants are telling the truth, can
insert questions to act as lie detectors)
§
Face validity( concerns the issue of whether a self report measure looks
like it is measuring what it is supposed to be measuring)
§
Concurrent validity ( comparing performance on a new self report measure compare
scores to demonstrate concurrent validity)
§
Predictive validity(predict outcomes you can assess the predictive validity of
a measurement)
§
Constructive validity (looking at underlying construct of a test e.g. should
represent theoretical views
Observations
§
Naturalistic observation
(everything left how it normally is = high ecological validity)
§
Controlled observations (variables
controlled by researcher = low ecological validity)
§
Unstructured observational technique (researcher notes all
behaviour used with unpredictable situations. – may not pick up on more subtle
details)
§
Structured observational technique (also called
systematic observations, researcher uses various systems to organise
observations such as: research aims
deciding on area to study, observational
systems how do you record all the behaviour, sampling procedures who you observe and when.
§
Participant and non participant (observer
may be participant in study too likely to affect objectivity)
§
Overt and covert (overt is when participants are aware
they are being observed this may reduce validity as they change their
behaviour)
Designing observational
studies
§
Recording data need
operationalisation -divide the behaviour into a set of component behaviours.
This is called a list of behavioural
categories or behaviour checklist
§
Sometimes each behaviour
is given a code to make recording easier = coding system
§
A further method is to
provide a list of behaviours and characteristics and use a rating scale
Behavioural
categories should:
§
Be clearly operationalised and
objective. Observer shouldn’t be able to make inferences
§
Cover all component behaviours and
avoid waste basket category
§
Be mutually exclusive, should
only mark one category at a time
Sampling procedures
§
Event sampling (counting number of
times behaviour occurs)
§
Time sampling( recording behaviours in
a time period)
Evaluation of
Observational method
+what
people say is often different to what they say they do and observations are
more valid than self report techniques
+naturalistic
observations give realistic picture and have high ecological validity
+observational
research provides a means of conducting preliminary investigations this
produces a hypothesis for future research
-little
control of extraneous variables in naturalistic observation
-observer
may see what they expect to see = observer bias
-if
observers don’t know they are being observed it can cause ethical issues
- If
they do know they are being observed they may change their behaviour
Content Analysis
Observations can be made directly
(observe first hand) or indirectly (through tv or magazine.) Content analysis
is the analysis of content of something. It is a form of indirect observation.
§
Sampling method (e.g.
time or event sampling)
§
Method of recording
(behavioural categories or coding system)
§
Method of representing data
(quantitive or qualitive data)
+ high ecological validity as it is
based on observations of what people actually do +sources can retained or accessed by
others so it is replicable and therefore tested for reliability
-Observer bias
can reduce objectivity and validity of findings
-
Likely to be culturally biased as interpretation will be affected by
language and culture of observer.
Correlational
analysis: inferential statistic tests
WILCOCON
T TEST This test is used for tests of difference where pairs of
date are related, such as when repeated measures design has been used. A matched pairs is also a
related design as participants have been matched.
WHEN TO USE = the hypothesis states a
difference between 2 sets of data. Two sets of data are pairs of scores from
one person, or a matched pair. Data has to be ordinal or interval.
|
One tailed
|
0.05
|
N is the degree
of freedom. You have 15 participants, using a 2 tailed study and looking
at a significance level of 2% the critical value is 25. Observed value is
19. To be significant the observed value must be smaller than or equal to
critical value.
|
|
0.01
|
|
Two tailed
|
0.10
|
0.02
|
|
N
|
|
|
|
9
|
8
|
5
|
|
13
|
21
|
17
|
|
15
|
30
|
25
|
|
19
|
53
|
46
|
Mann-Whitney U test
This is a test of difference. It
enables us to test if there is a difference between two sets of data. Tests of
difference are usually used for experiments, for example seeing if noisy conditions
reduce the effectiveness of revision.
WHEN TO USE:
·
If the hypothesis states
a difference between two sets of data
·
Two sets of data are
from separate groups (INDEPENDENT
GROUPS DESIGN)
·
The data is ordinal or interval
|
|
4
|
6
|
8
|
10
|
|
4
|
1
|
3
|
5
|
7
|
|
6
|
3
|
7
|
10
|
14
|
|
8
|
5
|
10
|
15
|
20
|
|
10
|
7
|
14
|
20
|
27
|
|
You have 4 participants in group 1 and 6 in group 2 the
critical value is 3. To be significant the observed value must be equal or less
than the critical value. If the observed value is 9.24 the result is not
significant. |
Spearman’s
Rho (spearman’s
correlation test)
This
test is used to determine whether a correlation between 2 variables is
significant or not. A perfect positive correlation is +1; a perfect negative is
-1.0. A figure of 0 = no correlation. In
this test as the no. of ppts increases the number needed for significance
decreases. E.G. Rahe et al found a positive correlation of +118 due to the number
of ppt (2,700) the correlation is significant.
WHEN TO USE =
·
If the hypothesis states a correlation between the two co
variables
·
If the two sets of data are paired scores (if they are related)
·
If the data is ordinal or interval
There is no requirement to learn any
calculation. You need to know how to determine if a correlation is coefficient
is significant. If the observed value is 0.58 and you have 30ppt using a one
tailed test we can say it is significant at p<0.05 as 0.58 is >0.306
WHY WE USE IT =
·
State the alternative and null hypothesis
·
Record the data and rank each co variable and calculate the
difference
·
Find observed value of rho (correlation co efficient)
·
Find critical value of rho
·
State the conclusion
Chi-Squared (x2) Test
A chi squared test is used to test the significance of nominal data.
We use this test when we have counted how many occurrences there are in each
category (frequency data.)
WHEN TO USE =
·
If the hypothesis states a difference between two conditions or an
association between co-variables
·
The sets of data must be independent
·
Data must be in frequencies (nominal) frequencies must not be
percentages
There
is no requirement to learn calculation. You need to know how to determine if an
observed value is significant. If x2=1.984 and there was 4 ppt using a 2 tailed
test we can say that the test is not significant as 1.984<9.49
4 levels of data (NOIR) =
1.
Nominal (names)
2.
Ordinal (order)
3.
Interval (gaps)
4.
Ratio (has a true 0)
3types of experimental design =
1.
Repeated measures design (same ppt do
each condition)
2.
Independent group design (different ppt
per condition)
3.
Matched pairs design (ppt matched
together according to variables)
Inferential
statistics help us draw inferences (conclusions) from the data tested.
Different inferential tests are used depending on the level of data used and
the experimental design used.
To find the
critical value you need:
·
The
degree of freedom (number of participants in study)
·
One
or two tailed test
·
Significance
level
·
Whether
the observed value needs to be < or > the critical value, to be
significant
EXAMPLE:
There are 20 people in your study. You used a directional
(1tailed) hypothesis. 5% significance level and you have an observed value of
53. T< tells us that it needs to be
bigger than. 53 is smaller than 60 which is the critical value. 0.05 is the
significance level.
|
One tailed test
|
0.05
|
0.01
|
|
Two tailed test
|
0.10
|
0.02
|
|
N
|
T = 53 T<
|
|
|
19
|
53
|
46
|
|
20
|
60
|
52
|
|
21
|
67
|
58
|
|
22
|
75
|
65
|
The
scientific process
Induction =
reasoning from particular to general. Make observations create testable
hypothesis, conduct a study, draw conclusions then propose a theory.
Deduction = reasoning from general to
particular. Make observations, propose theory, generate hypothesis based on
theory, theory tested and conclusions drawn.
Peer Review = the assessment of scientific work by others who are experts
in the field. The intention of peer review is to ensure it is published to a
high quality.
3 main purposes:
1.
Allocation of research funding (funded by governing bodies or charities it’s in their
interest to ensure they are not wasting money)
2.
Publication (ensures that
research is correct and not fraudulent)
3.
Standard deviation = measure
of the spread of data (don’t need to work it out in an exam)
+more precise as lots of
data taken into account. - Affected by extreme values
Range = difference between
smallest and largest.
+easy to calculate
-
Affected by
extreme values
|
|
Assessing
research rates of universities ( funding can
depend on if peer review ratings are good)
Issues with peer review=
·
Hard to find expert
·
Rival researchers
·
Publication bias
·
Preserving status quo
Biological + behavioural = psychology
is a science
Psychodynamic + cognitive = psychology
is not a science
Based on empirical, falsifiability,
replicable, objective, theory construction and hypothesis
LAYOUT FOR PSYCHOLOGICAL REPORT:
1.
Title
2.
Abstract (summary of
everything)
3.
Introduction (literature
review, aims and hypothesis)
4.
Method (procedure,
design, sampling)
5.
Results ( visual
representation of data eg chart significance of results)
6.
Discussion (conclusion
relate back. Supporting refuting evidence, flaws in study, extraneous
variables)
7.
References (sources /
bibliography)
8.
Appendices
Histograms = continuous data, interval or
ratio. No gaps between bars
Bar
charts = compare nominal data with categories
Scatter
graph = 2 sets of data looks at correlation. Can only have 2 variables Ordinal
or ratio or interval but not nominal.
Frequency/
line graph = continuous data. Individual points plotted no grouping.
Tables = use mean and
standard deviation.
Correlational Analysis
A
correlation is a relationship between two variables. If two variables increase together
then they are co-variables. There can also be zero correlation if nothing is
happening. Correlation is not a research method, it’s a method of analysing
research data.
Scatter grams
A correlation can be illustrated using a scatter gram.
For each individual we obtain a score for each co-variable. The co-variables
determine the X and Y position of each dot. In other words for one co-variable
you locate its position on the X axis (horizontal) and for the other
co-variable you locate its position on the x axis (vertical.)
Correlation Coefficient
The
scatter of the dots indicates the degree of correlation between co-variables.
If the dots are closely grouped together roughly forming a line from bottom
left to top right this indicates a positive correlation. The closeness of a correlation
is described using a correlation co-efficient. A correlation
co-efficient is a number, and it has a maximum value of 1.0. +1.0 is a perfect
positive correlation and -1.0 is a perfect negative correlation.
Some
correlation coefficients are written as -.76, whereas others are +.76. The plus
or minus sign shows whether it has a positive or negative correlation.
The
coefficient number tells us how closely the co variables are related. -.76 is
just as closely related as +.76, it’s just that -.76 means that as one variable
increases the other decreases.
A correlational hypothesis
In
a study using correlational analysis
there is no independent or dependent variable. When conducting a study
using correlational analysis you need to produce a correlational hypothesis
that states the expected relationship between co-variables.
Linear and curvilinear
If the relationship is not linear it
is curved then it is called a curvilinear correlation. For example anxiety and
performance do not have a linear relationship.
Performance on many tasks is lowered when anxiety is high.
+ Can be used when it
would be unethical or impractical to manipulate variables and can make use of
existing data
+ If a correlation is
significant then further investigation is justified. If correlation is not
significant then you can rule out a simple linear relationship.
+ As with experiments
the procedures can be repeated again this means that the findings can be
confirmed.
-
Cannot
demonstrate cause and effect
-
People
misinterpret correlations, this can lead to public misunderstandings
-
There may be
intervening variables that can explain why the co-variables are linked.
-
Using
Correlational analysis means the study lacks internal and external validity
How to
know if it is an Investigation = it does use a questionnaire to collect data,
it’s not an experiment as they have an IV and DV, and it’s not a case study or
observation.
Cross Cultural Studies
This is a way of seeing if cultural practices affect
behaviour. It is a kind of natural experiment
+ This
technique does enable psychologists to see whether some behaviours are
universal.
-
Observer
bias can be a problem because researchers have expectations about how the study
should go and this could affect their measurements. The use of local
researchers can help overcome this
-
Could
be communication difficulties which can be overcome by using indigenous (local) researchers
-
Researchers
may use test or procedures developed in the US this may make other cultures
seem abnormal. This is called imposed
ethic.
-
Group
of participants selected for study may not be representative of that culture
and we may make generalisations.
Meta analysis
This is when the results of many studies that have a similar
hypothesis are combined. The IVs tend to be measured in different ways so the
researcher uses ‘’effect size’’ as
the DV in order to asses overall trends.
+ analysing
the results from a group of studies rather than just one study means that the
conclusions are more reliable
-
Studies
are not truly comparable as they use such different research designs.
The
Multi- method approach = Very few studies use one method. A combination of all sorts
of methods is called the multi method approach.
Role Play
In some investigations participants are required to take on
a certain role and their behaviours can be observed as if they were everyday life.
+ This
enables researchers to study behaviour which might otherwise be impractical or
unethical to observe.
-
Role
play is acting and so peoples actual behaviour can be questioned as role play
is only a prediction.
Longitudinal and cross-sectional studies
Longitudinal study is conducted over a long period of time
in order to observe long term effects. In a cross sectional study one group of
participants are young and one are old and they are compared for example.
+ Longitudinal
studies have high control over participant variables
+ Cross
sectional studies have the advantage of being quick.
-
In a longitudinal study there is the
problem of attrition (study getting
smaller because people drop out) a certain type of person may drop out and this
may leave the study bias
-
In
longitudinal study participants are likely to become aware of the studies aims
and this may affect their behaviour
-
Longitudinal
studies take a long time to complete and take lots of finance
-
In
a Cross sectional study the participants may vary in more ways than just the
behaviour being researched
Cohort
effects occur because a group (or cohort) of people who are all the same
age share certain experiences
·
In longitudinal studies they may only consider one
cohort and that isn’t generalisable because of the unique characteristics of
the cohort.
·
In a cross sectional study may also suffer from
cohort effects.
PSYCHOLOGY
RESEARCH METHODS (EXPERIMENTAL METHOD):
Experimental hypothesis = prediction of what we think is
going to happen
Null hypothesis = if there is no correlation and no relationship
Alternative hypothesis = something is happening in our
experiment
Directional (one tailed) hypothesis = states that one thing is going to
happen
Non directional (2 tailed) hypothesis = states that a few things could happen
Probability
= measure of how likely something is going to happen. Can never be 100% sure,
use probability of 95% to express degree of uncertainty. This = p = 0.05.
90% sure would have a P value of 0.1.. 5% p level is compromise between being
lenient and stringent. In some studies
you need to be really certain, so probability p=0.01 or even 0.001
Value
of p = SIGNIFICANCE LEVEL
"less than or equal to" sign: ≤
The "greater than or equal to" sign: ≥
|
|
|
TRUTH
|
TRUTH
|
|
|
|
DRUG WORKS
|
DRUG DOES NOT WORK
|
|
TEST RESULT
|
DRUG WORKS
|
TRUE POSITIVE
|
FALSE POSITIVE
|
|
TEST RESULT
|
DRUG DOES NOT WORK
|
FALSE NEGATIVE
|
TRUE NEGATIVE
|
Type 1
error = we think something is going on, but there is not. (P VALUE TOO
LEIENT) Type 2 error = we think
nothing is going on, but there is.(P VALUE TOO STRINGENT)
LAB EXPERIMENTS conducted
in controlled environment
+ can infer cause and effect
+the experiment is replicable and
therefore reliable
-
Demand characteristics,
investigator effects
-
Reduced internal validly
as it is not like the natural setting
-
Low ecological validity
FIELD EXPERIMENT
natural environment
+high ecological validity, experiment
is replicable
+less experimenter effects because
participants are less aware that they are in a study
-
IV harder to control
-
Lower external validity
because it’s in natural environment so there are other outside factors
NATRUAL EXPERIMENT
makes use of existing IVs, iv not manipulated
+high ecological validity as it is
peoples real experiences
- Participants not
randomly selected this reduces validity
- Study is not
easily replicated
- Extraneous
variables hard to control
Internal
reliability = measure if something is consistent within its self
Internal
reliability assessed by:
·
Use the split half method to compare a person’s
performance. There should be a close correlation as a measure of high internal
reliability
·
Face validity = does it look right?
·
Concument validity = compare performance
Improve
internal reliability = select test items that have the most similarity remove
certain items to make the relationship stronger
External reliability = a measure of consistency over
several different occasions outcome should always be the same
External reliability
asses =
test retest method. Leave it long enough so don’t forget what was said the
first time. Give same test to save participant different occasions
External reliability
improved =
depends on experimental method used. It could be that the test questions are
ambiguous or it could be that the interviewer needs more training
Internal
validity = concerns what goes on inside study whether researcher did what they
intended to do
Factors
affect internal validity =
·
Situational variables e.g. time of day, order effects
·
participant variables e.g. age
·
participant effects e.g. demand characteristics
·
experimenter bias e.g. can be direct or indirect
·
individual differences e.g. temperature
External
validity = things outside of study, the extent to which the results of the
study can be generalised to other situations and other people
3
types of external validity =
1. ECOLOGICAL VALIDITY
2. POPULATION VALIDITY
3. HISTORICAL VALIDITY
External
validity improved = enhanced through replication
Mundane
realism = when research replicates everyday life
Ecological
validity = if we can apply it to the real world
|
Design
|
strengths
|
limitations
|
|
REPEATED MEASURES DESIGN
|
easy to compare results, no problem with
participant variables, save time and money as don’t need as many participants
|
Order of conditions may
affect performance, may have practice effects. People may do better in
condition2, can deal with this by counter balancing – split into 2 groups
both do each condition
|
|
MATCHED PAIRS
DESIGN
|
Controls participant
variables
|
Time consuming and
difficult to match participants, should always do pilot study
|
|
INDEPENDENT GROUPS
DESIGN
|
No issue with order
effects
No practice effects
|
No control of participant
variables eg age intelligence
|
|
Ethical issue
|
How is it managed
|
|
Protection for physical and psychological harm
|
Reassure participants that they can stop the study if
they are distressed. Also have someone observing for people being harmed
|
|
Informed consent
|
Sign waver beforehand about what is going to happen. Or
if under 18 get parent or guardians permission
|
|
Privacy issue
|
Don’t release names or details, also protect human rights
|
|
deception
|
Full de brief before the study letting them know what
will happen. Or you can get presumed consent
|
|
Right to withdraw
|
Tell participant before study that can withdraw from the
study at any point even after it has finished
|
|
Confidentiality
|
tell participants that no personal details will be used
if results are published – sign agreement before hand letting them know
|
