psychology as science Flashcards
history of knwoledge acquisition
informatio centralised in libraries, hard to access
limited public access
information filtered through academic experts
now information on internet its widely academic
widepsread dissemination of misinformation
consequences of internet
need to be critical consumer evaluate information myself consult credible, reliable information services use primary sources experts publish peer reviewed journals
HMR hoax
1998 wakefield claimed children who had mmr vaccine developed autism
ignore high quality research showing no link
biased and uninformed media coverage
measles outbreak due to parents not getting vaccinated out of fear
how to obtain knowledge
authority figures:celebs, religious leaders, political leaders, consultants and senior academics
introspections and intuition: psychology is common sense
experience: personal or second hand
scientific method: key elements, objective measurements, refutation, reinterpretation
authority figures
abuse status to exploit
gwyneth paltrow fined $145,000 for unproven health claims
introspection and intuition
humans lack insight about how they function human reasoning is biased irrational judegments influenced by prejudices confirmation bias availability heuristic
illusory pattern detection
prone to seeing patterns in randomness eg faces in clouds
scientific method
theory generates hypothesis hypothesis leads to testable prediction obtain data that's used to test hypothesis interpret data used to create conclusion relate findings to hypothesis generate more hypotheses
characteristics of scientific method
not defined by subject matter
not defined by use of experimental method
not defined by obtaining qualitative data
defined by approach-rational, systematic, objective, careful
results open to scrutiny by skeptical others, potentially falsifiable, ideally reproducable
difference between science and pseudo science
science produces testable claims that are open to disproof
scientific claims must acknowledge all findings whether supportive or not
pseudo science uses evidence selectively to support belief ignores everything contradictory evidence
limitation to scientific method
restricted to testable questions not claims eg ‘god exists, but doesn’t reveal himself’
theory, hypothesis, predictions and data
theory: evidence based conceptual framework that tried to explain set of facts and observations, used to make testable predictions
hypothesis: proposed explanation derived from theory
prediction: scientific, testable prediction stemming from hypothesis
data: facts attempting to explain prove or disprove a theory
why do we quantify things
defining principle of science is measurement as can be objectively obtained
measurements must be reliable and valid
levels of measurement
nominal: numbers used as names, count how often each number occurs, frequencies of categories
ordinal: numbers used as ranks, attitude scales
interval: scale has equal intervals between points but no true zero point, IQ score, temperature
ratio: measurements made on scale with equal intervals and true zero point, reaction times/error scores
reliabilty
results msut be consistent and reproducible
a score = true score + error
error due to:natural performance variation (with/between variation), imprecision in defining and measuring psychological constructs (exactly defining aggression)
meaures of reliability
test retest
alternate forms
aplit half
inter scorer
factors affecting reliability
phenomenon itself (traits vs states) precision of measurement sample size (bigger > small) time between tests (shorter > longer) variability in performance (high > low) format of test (multiple choice > true/false) between individuals variability in scores (high > low)
replication crisis
causes: small sample sizes encourage flukes
straight replications are rare
file drawer problem: hard to get failed replications published
solutions: replications, meta analysis, converging operations
Nosek 2015
replicated 100 experimental and correlational studies from 3 prestigous journals
97% original studies had significant results
36% replications had significant results
combining original and replication results left 68% studies had statistically significant effects
validity
measure what its supposed to be measuring
measure can reliable but not valid
reliable but invalid: phrenology
gall and spurzheim
different parts of brain responsible for different mental faculties
highly developed faculties led to larger brain regions
larger brain regions reflected by bumps on skull
reliable as scientific but invalid as bumps no relation to brain
reliable but invalid: brain size
paul broca 1870s
292 male brains, 140 female brains
‘women on average less intelligent than men, small size brain depends on intellectual inferiority
measures of validity
face content criterion construct ecological/external
factors affecting validity
only influence on dependent variable is manipulation of independent variable
norms and standardisation
stratified random sampling
control group to compare against
ecological validity
to what extent is results generalisable to real world
Experimental method
Best for identifying causal relationship
X causes Y if X occurs before Y, Y doesn’t occur in absence of X
Good experimental designs ….
Maximise validity
Internal: ensure dv changes due to manipulation of iv
External: generalise from participants to other groups
Threats to internal validity
Time: history, maturation, selection-maturation interaction, recreated testing, instrument change
Group: initial non-equivalence of groups, regression to mean differential mortality, control group aware of status
Participant reactivity threats: experimenter effects, reactivity, evaluation apprehension
How’s validity affected by history
Extraneous events between pre-test and post-test affect participants performance in post-test
Solution: add control group
How’s validity affected by maturation
Participants may change during course if study eg get older or fatigued
Solution: control group
How’s validity affected by selection-maturation interaction
Different participant groups have different maturation rates, affect how participants respond to experimenters manipulation
Solution: ensure groups only differ on one independent variable
How does repeated testing affect validity
Taking pre-test May alter result some post-test
Solution: avoid repeated testing or add control group who don’t complete pre-test
How does instrument change affect validity
Eg experimenter tests all of one group before testing another, become more practiced/bored while running study
Now two systematic differences
Solution: use highly standardised procedure random allocation to conditions, familiarisation with behaviours before observations
How does selection affect validity
Cohort effect: groups differ on many variables eg gender
Can’t conclude observed differences solely due to independent variable
Solution: matched group design
How does regression to mean affect validity
Participant who score very low or very high on one occasion tend to give less extreme scores on another occasion
Solution: random selection, avoid floor and ceiling effects with scores
How does differential mortality affect validity
Subject attrition, sample no longer representative
Solution: difficult to fix
How does reactivity affect validity
Hawthorne effect: increase in productivity due to awareness of being observed
Draper 2006 review: productivity affected by: material factors, motivation, learning, feedback on performance, attention/expectation of observers
Implications:act of measurement can affect very thing being measured
Experimenter effects on validity
Expectations affect performance
Pygmalion effect-teachers affect IQ of pupil
Placebo effect- drug expectation affect drug effects
Solution: double blind
Quasi experiments
No control over allocation or timings of manipulations of iv
One group post test design: prone to time effect, no control group
On group pre/post test: prone to time effects, baseline to compare
Interrupted time series: measure at periodic stages, prone to time effects
Static group comparison: no random allocation, observe differences not solely due to iv
True experiment designs
Post test control group: random allocation
Pre/post test control group: random allocation, groups comparable before and after manipulation
Solomon four group: two groups pre/post test control, two groups post test control group, ensure pre test not affect performance
Between groups vs within subjects
Between (independent measures): each subject participates in only one condition
Within (repeated measures): each subject does all conditions
Mixed designs: mix of both
Advantage/disadvantage of between groups
Straight forward Needs more subjects No carry over effects between conditions Lower sensitivity to experimental effects Reversibility of conditions unimportant
Ada vantages/disadvantages of within subjects
Complicated Fewer subjects Possibility of carry over effects Higher sensitivity to experimental effects Reversibility of conditions essential
Cross sectional vs longitudinal
Cross sectional: different groups for each time phase of study
Longitudinal: each participant is measured repeatedly over time
Within subjects and order effects
Oder effects: boredom, practice, fatigue
Randomise order of conditions to eliminate impact of order effects
Disadvantage of experimental method
Intrusive: participants know being observed may affect their behaviour
Experimenter effects
Not all phenomenon can be experiments for practical/ethical reasons , some phenomenon only investigated as quasi experimenters
Why do we need ethical guidelines
Belmont report: respect for persons, beneficence, justice
psychology specific codes of practice
british psychological society
american psychological association
BPs code of ethics
- regularly review documents
- record decisions regarding ethical issues
- ethical principles: respect(privacy, consent), competence(professional standards), responsibility(respect welfare), integrity(honest, unbiased)
- legal obligations: health and care council registration, competence, indemnity insurance, disclosure and barring service checks, equality act, data protection, freedom of information act, safeguard children, mental capacity act, mental health act
APA code of conduct
- beneficence and nonmaleficence(benefit those they work with and do no harm)
- fidelity and responsibility(establish trust and be aware of professional and scientific duties
- integrity(promote accuracy and honesty in science, teaching and practice)
- justice(exercise fairness and ensure equal opportunity to benefits
- respect for peoples rights, dignity(respect worth of people, privacy, confidentiality
key points on human research
risks explained participation voluntary valid informed consent advice given confidentiality maintained deception-if leads to harm its inappropiate debrief after study
informed consent requires:
- voluntary participation
- consent-capacity to make/communicate decision, understand the information, weigh up consequences logically
- inform participants of purpose of research, duration, procedures, right to withdraw once started, factors affecting willingness to participate, research benefits, incentives
informed consent with special groups
- problem of groups who cant give consent themselves eg children or demented
- obtain informed consent from carers
- where procedures involve risk of harm, obtain informed consent from individual and consult ethic committee
- child’s avoidance of testing should be taken as withdrawal from study
informed consent and power relationships
be aware prisoners/institutionalised individuals and students may feel obliged to say yes
belmont report prohibits coersion
benefits of informed consent
force researchers to think more about theri research
encourages trust and better rapport with participants
better recruitment rates
costs of informed consent
‘delays, bureaucracy’
‘middle class’ attitudes to informed consent alienate or confuse other social groups and ethnic minorities
some vulnerable groups in research-able as obtaining consent difficult
hawthorne affect-behaviour altered as participants aware of being observed
inducements to participants
not to be excessive
not to coerce participation in risky situations
use of deception
only used in unavoidable
precludes informed consent
makes people distrustful of psychologists
consider participants reaction to finding out been misled
debrief participants as soon as
consult ethic committee
deception varies in extent:
-informed consent but not knowing condition allocated to
-consent but not know full details until after study
-involved with no prior knowledge or consent
-consent to study but misled to true aim
Film and voice recording
experiments/therapy sessions: make/use only with participants knowledge and consent
observational/naturalistic studies: no knowledge or consent needed unless individuals are identifiable or harmed
boundaries of competence
important in applied areas
must have appropriate skills and expertise
keep knowledge up to date
acknowledgement of limitations and boundaries when dealing with non-specialists
important for all psychological research
sharing data
maintain confidentiality
remember informed consent
avoid plagiarism
participants can remove data from data set
collecting participant data
remain confidentiality
only acquire and retain personal information that’s necessary
participants have right to remove data
debriefing
full explanation of what participant is involved in
avoid evaluative statements
consideer effects of self esteem
provide contact details for follow up questions
dont justify unethical/misleading experiments
if psychological/physical problems arise, researcher must alert participants and refer to expert for treatment if needed
ethical issues in internet research
internet surveys or observation studies
need to distinguish between internet chat rooms, private email and correspondence and instant messaging
lack of interactivity poses issues: difficult to ensure informed consent, difficult to ensure adequate debrief, need to ensure confidentiality, widens access to study participants
stages in scientific investigation
obtain data, from sample taken from population
descriptive statistics, reveal info lurking in data
inferential statistics, use data from sample to reveal characteristics of population from which sample data selected
descriptive statistics
summary statistics, means, medians, modes, describe typical performance
frequency distribution, describe prevalence of different types of performance
quantitative, frequency of scores of single variable
qualitative, frequency for mutually exclusive categories
relative frequency distribution
comparing groups with different tools
rf= (cell total/overall total) x 100
raw frequency and relative frequency
graphs have same pattern, but different scales
normal distribution
mathematical abstraction which describes many frequency distributions in real life
properties of normal distribution
bell shaped
asymptotic extremes
symmetrical around means
mean, median, mode have same value
skewed distribution
lack symmetry as mean median mode different values
positively skewed-highest point to left of mean
negatively skewed-highest point to right of mean
skewed data distorts perception of mean
solutions: use median to describe data, principled treatment of outliers
kurtosis
measure of tail heaviness
mesokurtic distribution-like normal distribution
positive/high kurtosis-fatter tails, more out outliers
negative/low kurtosis-thinner tails, fewer outliers
type of statistics
descriptive-quantitative description of data, means, median, mode
inferential-help decide whether or not any observed patterns in our data have occurred merely by chance
summary descriptives
measures of central tendency
measures of dispersion
measures of central tendency
mean
median
mode
mode
most frequent score in set of scores
advantage: simple to calculate, only used with nominal data
disadvantage: may be unrepresentative, misleading, more than one mode in set of scores
mean
add all scores together/total number scores
advantage: uses information from every single score, resistant to sampling fluctuation
disadvantage: susceptible to distortion from extreme scores
median
scores arranged in size, median is either the middle score or average of middle two scores
advantage: resistant to distorting effects of extreme high or low scores
disadvantages: ignore scores numerical value wasteful data, more susceptible to sampling fluctuations than the mean
measures of dispersion
range
standard deviation
range
difference between highest and lowest scores
advantage: quick and easy to calculate, easy to understand
disadvantage: unduly influenced by extreme scores, convey no information about spread of scores between highest and lowest scores
standard deviation
average difference of scores from the mean, bigger the SD the more scores differ from the mean and between themselves, less satisfactory the mean becomes as summary of data
advantages: uses information from every score
disadvantage: not intuitively easy to understand
complications of mean and SD
obtain mean and SD from sample, very rarely from parent population
sometimes content to describe sample usually want to extrapolate to population
sample mean is good estimate of population mean
sample SD tends to underestimate population mean
when using sample SD as estimate of population SD divide by (n-1)
when using sample SD as description of sample divide by n
normal distribution and standard deviation
all normal curves share these properties
68% scores in range of mean +/- 1 SD
95% scores in range of mean +/- 2 SD
99.7% scores in range of mean +/- 3 SD
standard error of mean
SD of set of sample means
how much variation within set of sample means
standard error = SD/ square root of n
if SE is small, obtained sample means more likely to be similar to true population mean
increasing sample size reduces size of SE
error bars show mean +/- 1 SD of mean
normal distribution
- mathematical abstraction whihc conveniently describes many frequency distributions of scores in real life
- area under curve directly proportional to relative frequency of distribution
- area under curve directly proportional to probabilities of observations
- probabilities expressed as p values between 0 and 1
relationship between normal distribution and standard deviation
- standard deviation cuts off constant proportion of distribution of scores
- 3 standard deviations on either side of mean
z-scores
standard scores
- states position of raw scores in relation to mean distribution, using standard deviation as unit of measure
- Z = raw score - mean / standard deviation
raw score distribution and Z-score distribution
raw score - X expressed in original units of measure
z score - X expressed in terms of its deviation from mean
why use Z scores
easier to compare scores from distributions using different scales
-enable us to determine relationship between on score and rest of scores
logic of statistical tests
scores normally distributed around mean
sample means normally distributed around population mean
-differences between sample means are normally distributed around zero
central limit theorem
- sample means normally distributed around population mean, regardless if actual shape of population itself
- any given sample mean can be expressed in terms of how much it differs from population mean
- deviation from mean is same as probability of occurrence
Type 1 and Type 2 errors
type 1: FALSE POSITIVE, falsely reject null as believe experimental manipulation had effect when it didn’t
- type 2: FALSE NEGATIVE, falsely retain null hypothesis as believe experimental hypothesis has not had an effect when it has
- any observed differences between two sample means could in principle be either ‘real’ or due to chance, never tell for certain
- larger the difference, less unlikely to be by chance
0.05 significance level
- set probability of making type 1 error at 0.05
- accept difference between two samples as ‘real’ if difference of size likely to occur by chance, 5% of time
summary of typical experimental procedure and analysis
- perform experiment, find mean of each sample and difference between means
- assume null hypothesis and two samples are still from same population
- assess probability of obtaining by chance a difference between sample means
- if probability of 0.05 or smaller reject null hypothesis, if bigger than 0.05 then accept null hypothesis
