Methods and stats Flashcards
what are the four levels of measurement
N - nominal
O - ordinal
I - interval
R - ratio
Nominal
- what type of data
- what type of stats
qualitative data
categorical
named
mode as would tell us which of the categories is the most commonly occurring
Ordinal
- what type of data
- what type of stats
quantitative
numbers have an ordered relationship
numbers indicate position on a list
eg first, second, third
differences between adjacent scores do not represent equal quantities
the median, range and interquartile range are appropriate as these measurements are based on position
Interval and ratio
- what type of data
- what type of stats
quantitative
numbers say what they mean - numeric properties are literal
interval as no absolute zero
the mean, standard deviation and standard error are appropriate descriptive stats
ratio as there is an absolute zero
what is parametric data
- The scores (the DV) must be at an interval or ratio scale.
- The data must be normally distributed.
- The groups must have homogeneity of variance (similar variances).
what are non-parametric tests
tests that make few or no assumptions about the shapes underlying the population distribution
also known as distribution free tests
can be used on low level data such as ranked data
parametric tests often have a non-parametric equivalent
is this test parametric or non-parametric?
chi-squared
non-parametric
is this test parametric or non-parametric?
Spearman’s Rho
non-parametric
is this test parametric or non-parametric?
paired samples t-test
parametric
is this test parametric or non-parametric?
independent samples t-test
parametric
is this test parametric or non-parametric?
Wilcoxon T-test
non-parametric
is this test parametric or non-parametric?
Mann-whitney U test
non-parametric
is this test parametric or non-parametric?
Pearson’s correlation coefficient
parametric
pair up the seven tests with their parametric / non-parametric equivalents
chi-squared & n/a
Wilcoxon T-test & paired samples t-test
Mann-Whitney U test & independent samples t-test
Spearman’s Rho & Pearson’s correlation coefficient
advantages of non-parametric tests
analyses can be simplistic and easier to complete
useful for data on nominal or ordinal scales
can be used with small sample sizes
useful if data violates the assumption of normality
useful if data is severely skewed or has outliers
makes fewer assumption so tests can be more robust
disadvantages of non-parametric tests
less powerful, typically only make use of ordinal information only. as such large sample sizes may be needed to find significance
due to having to assign ranks, analyses can be annoying if sample size is very large
utilitarianism
what who
where/ how is it in practice
but…..
the greatest happiness principles
Jeremy Bentham (1748-1832)
John Stuart Mill (1806-1873)
hedonic calculus
constrained utilitarianism - human and animal research based on cost-benefit analysis, but with absolute limits defining acceptable practices
but what are the limits of acceptability? who decides?
who are the professional codes of conduct for psych
American psychological association
british psychological society
society for neuroscience
world medical association (including BMA) - the declaration of Helsinki
examples of why we should worry about human research
military studies in Nazi Germany Tulane - heath curing homosexuality tuskgee syphilis study zimbardo milgram and asch social conformity study the 'unfortunate experiment' NZ
ethics in human research - key points (8)
informed consent motivation for being a subject degree of risk or personal harm right to withdraw confidentiality - data protection act protection of participants debriefing follow-up procedures to detect and mitigate any lasting adverse effects
who reviews our ethics
at university - school ethics committee and UTREC (university teaching and research ethics committee)
ethics of research on animals - why worry? key points (5)
invasiveness behaviour manipulations field work housing conditions genetic manipulations
viewpoints on animal research and their weaknesses
absolute anti-research
ethical status of animals = humans
weakness
unreasonable conclusions
removes limits to ‘direct action’, actually requires it
failure to recognise human awareness (Singer)
what is direct action
terrorism staff firebombed, kidnapped, threatened with murder
viewpoints on animal research and their weaknesses
absolute pro-research
animals as tools, objects
weakness
unreasonable conclusions - kill all dogs for one human? animal ownership rather than stewardship
failure to recognise inherent worth of animals
modelling paradox for human disease - to the extent animals are not like humans, then they are poor models. to the extent that they are like humans, then we should not use them
animal research in the UK
2007
3.2 million procedures
by comparison 10 million cats in the UK kill 300 million animals and 2.5 billion fish and animals are consumed in the UK
ethical conduct of research key point (5)
plagiarism intellectual honesty and data analysis determining authorship conflicts of interest relationship with the media
what does Wilcoxon T-test do
establish if there is a change from one condition to the next
Wilcoxon T-test
what type of data
non-parametric (is the equivalent of paired samples t-test)
data must be at the ordinal scale or higher
what does the Wilcoxon T-test tell us
gives us information about the direction and magnitude of the difference between pairs of scores
steps in carrying out Wilcoxon T-test
- First describe the two data sets by calculating the median or mean (this depends on the type of data you have collected - for ordinal data use the median and for interval or ratio data use the mean)
- Calculate the difference between the two conditions (D = X1 – X2).
- Assign ranks to these differences ignoring the plus and minus signs and omitting any pairs of scores with a difference score of zero.
- Add back in the plus and minus signs to each rank. Add up the ranks with plus signs and then add up the ranks with minus signs. The smaller of these two values is your T value.
- Look up the critical value in the table (p.66) according to your N (omitting pairs of scores with a difference score of zero). If your T value is EQUAL TO OR LESS THAN the critical value you have significance.
- Report your Wilcoxon statistic as: T(n) = X, n.s. OR T(n) = X, p<0.05
what are the two types of Chi-squared
goodness of fit - deals with nominal data (so data is categorized)
test of association - two or more groups
Chi-squared goodness of fit - info
examines one nominal variable and looks at how participants are allocated to different categories within that variable
deals with frequency of scores in each category
categories must be mutually exclusive - independent
we are looking at the distribution of observed frequencies compared to expected frequencies
how good is the fit between the observed and expected frequencies
there are two possible null hypotheses
Chi-square goodness of fit hypotheses
null hypotheses
in general: that there is no difference between the distribution of our observed frequencies and the expected frequencies
1 - the population is evenly distributed across the categories (as you would expect by chance). there is no preference for one category over another
2 - the proportions in each category do not differ from a comparison population that has published expected frequencies
alternative hypothesis - there is a difference between the distribution of our observed frequencies and the expected frequencies - no specific prediction so two-tailed
how to carry out Chi-squared goodness of fit
draw a contingency table (table with categories, observed and expected frequencies)
1 - calculate expected frequencies
2 - record observed frequencies
3 - calculate Chi-squared
4 - check significance with degrees of freedom
5 - interpret results
degrees of freedom in Chi-squared goodness of fit
= number of categories - 1
when to use Chi-squared test of association
two nominal variables and want to know if they are associated
hypotheses in Chi-squared test of association
null - that two nominal variables are not associated (two nominal variables are independent
alternative hypothesis - that two nominal variables are associated
steps in how to do Chi-squared test of association
1 - create contingency table 2 - record observed frequencies 3 - calculate expected frequencies (row total x column total divided by total 4 - calculate chi-square 5 - check significance 6 - interpret results
degrees of freedom in chi-square test of association
(r-1)x(c-1)
what is reliability and 3 subsections
the extent to which a measurement is repeatable and consistent
- internal consistency
2 - inter-rater reliability
3 - test-retest
what is internal consistency
the extent to which tests or parts of tests asses the same characteristic, skill or quality
a measure of reliability of different test items intended to measure the same characteristic
internal consistency reflects the extent to which items of a test measure various aspects of the same characteristic and nothing else
what is inter-rater reliability
after observing the same behaviour, the recordings of two or more researchers should match
“calibrating” the observers
inter-rater reliability is dependent upon the ability of two or individuals to be consistent
training, education and monitoring skills can enhance inter-rater reliability
test-retest
consistency of a measure from one time to another
when an individual sits the same test multiple times, they should get the same result each time
the most popular indicator of reliability, quantified by correlation coefficient measurement
what is validity
3 subsections
the extent to which a test measures what it claims to measure
- construct validity
- external validity
- internal validity
construct validity
whether a scale measures the theorized psychological construct that it purports to measure
external validity
the degree to which the conclusions in your study would hold for other persons in other places and at other times
population validity - whether the sample represented the entire population
ecological validity - how well methods, materials and setting of the study approximate the real-life situation
internal validity
feeling confident you can make causal statements about what happened in your study
the independent variable causes the change in the dependent variable
your study is free of confounding variables
explain the trade off between internal and external validity
a tightly controlled experiment is likely to have strong internal validity but results will be harder to generalize
more naturalistic research will have better external validity, but will be poorer in terms of control and subsequent internal validity
name the eight threats to internal validity
selection maturation history repeated testing instrumentation experimental mortality selection-maturation experimenter bias
selection
differences between groups may exist pre-test that could interact with the independent variable and thus be ‘responsible’ for the observed outcome
aim to have unique individual characteristics evenly distributed between groups (such as sex, ethnicity, attitude etc)
maturation
participants may change in the course of the experiment or between repeated measure of the dependent variable
some of these changes are permanent such as growth whereas others such as fatigue are temporary
history
outside events may influence participants in the course of the experiment or between repeated measures of the dependent variable
often these are large scale events that affect participants’ attitudes and behaviours (eg war, natural disaster)
repeated testing
repeatedly measuring the participants may lead to bias. simply taking the pre-test may affect how participants do on the post-test
1 - participants may become less anxious; they may remember answers; become aware of concepts that weren’t originally known. subject to practice effects.
2 - features of the experimental situation may change from test session 1 to test session 2
instrumentation
a threat produced by changes in the measurement instrument itself such as changes in calibration of a mechanical measuring device or the proficiency of a human observer or interviewer
experimental mortality
between testing sessions participants may withdraw from the study
there could be differences between those who remain in the study and those who drop out
such as those who are less intelligent tend to drop out
selection-maturation
participant-related variables and time-related variables may interact
the difference between the groups is always present but only becomes evident through the maturation process
the participants selected into treatment groups may mature at different rates
experimenter bias
the experimenter intentionally or unintentionally biases the study and therefore influences the outcome
use a double blind study: an experimental procedure in which neither the participants nor the experimenter know the critical aspects of the experiment (who receives placebo and who receives treatment)
demand characteristics
a feature of the experimental situation that indicates to the participants:
how the researcher would ideally like them to respond in the study
the true purpose of the study
subsequently participants alter how they respond
participant reactivity
the influence the experimental situation may have on the participant
1 - negativistic participant role
2 - apprehensive participant role
3 - good participant role
blind design
the participant does not know which condition they are in for the study - reduces demand characteristics
double blind
neither the participant nor experimenters know who is in the experimental group and who is in the control group - removes both experimenter bias and demand characteristics
what is correlation
indicated an association between two variable. more specifically the strength and direction of a linear relationship between two variables
positive correlation
bottom left to top right
as one variable increases, so does the other
negative correlation
top left to bottom right
two variables moving in opposite direction to each other
no correlation
as one variable changes, the other variable does nothing
how do we visually show correlation
scatter graphs
the most important thing to remember about correlation
correlation does not imply causation
how to calculate correlation coefficient
• Correlation coefficient = r
• Correlation coefficient varies between -1 and +1
• Direction: sign of value shows ‘-‘ or ‘+’ relationship
• Magnitude: closer to 1 the stronger the relationship (ignoring the sign)
• If r=-1 or +1, a perfect linear relationship exists
• If r=0, there is no linear relationship
• Interpreting r
o O.7-1.00 = strong
o 0.5-0.69 = moderate
o 0.3-0.49 = weak
what are the two versions of correlation coefficient and why are there two
• Versions of the correlation coefficient
o Parametric: Pearson’s correlation coefficient
o Non-parametric: Spearman’s rho
how to carry out Pearson’s correlation coefficient
fill in table with x and y squared
plug into formula
how to carry out Spearman’s Rho
MUST FIRST RANK DATA
then same as pearson’s but with the ranks
degree choice and conservatism - do students studying different disciplines differ in their conservatism score background research
vocational choice may be a function of personality
we look for jobs that fufil our interests - Super 1957
vocational choice may reflect one’s own life style - Holland 1985
choosing a degree course may share similarities with occupation choice
the creation and evolution of the conservatism scale
conservatism - a desire to retain traditional social institutions
scale originally developed by Wilson and Patterson 1968 - 50 items
adapted and abbreviated by kirton 1978 - 30 items (scale investigates 3 dimension: religious, sexual and moral control)
abbreviated again Joe 1984 - 20 items
this latest version of the scale has not been standardised, scores from this scale may not be normally distributed
what is the Mann-Whitney U test trying to achieve
non-parametric test (ordinal data or higher) aimed at answering whether two groups perform differently on a task
how to carry out the Mann-Whitney U test
- Calculate the medians for each group.
- Rank scores (combining the groups). Zero values DO count.
- Total the ranks for each group separately.
- R= largest total of summed ranks. N1 = number of participants in the group that has the largest sum of ranks.
- Determine the U value using the equation: U = (N1 x N2) + N1 x (N1 + 1) – R 2 6. Look up the Mann-Whitney U test critical value on the table (p.67-68).
- N values in Significance Tables: smaller N value down column, larger N value across the row (if sample sizes are different).
- Table gives two ranges of scores. If your U value lies within one of these two ranges your result is significant. 9. U(N1,N2) = X, p<0.05 or U(N1,N2) = X, n.s.
dysfunction to neural circuits - name some of them and the psychiatric conditions underlying them
and separately some causes
Parkinson’s disease - motor circuits
Major depressive disorder - mood circuits
Alzheimer’s disease - cognitive circuits
damage can be caused by:
-damage to pathways, stroke
-loss of neural elements / populations, Parkinson’s
-disordered firing, epilepsy)
why would we carry out deep brain stimulation
there are a variety of disfunctions to neural circuits that underlie many symptoms of different psychiatric conditions
so successful treatment of these circuits has come about through stimulating particular circuits (called deep brain stimulation)
benefit was initially found in movement disorders, such as Parkinson’s
in patients that have had conventional therapies without success, DBS has been successful (loranzo and lipsman 2013)
how does DBS work
Electrodes implanted in specific region of the brain, subject is awake to aid placement
• Electrodes are connected to a programmable pulse generator that is implanted under the skin, below the collarbone
• Can change settings of frequency, pulse width, amplitude to maximise clinical improvement
• Mechanism of how DBS works is currently unknown
what are the problems with DBS - not ethical
with time and disease progression many symptoms in Parkinson’s disease develop even I those with DBS surgery
does not work for all patients
alleviates some symptoms but not a cure
problems with major surgery such as infection
possible side effects of mood disturbance
possibilities for the future with DBS
All actions, feelings, thoughts generated in the brain
• Therefore, those with actions, feelings, thoughts that are unacceptable to society (e.g. a pedophile, psychopath) have abnormal brains in some way
• These circuits could be silenced/re-programmed using DBS
treatments for spinal cord injury
worldwide, over 130000 people each year survive spinal cord injury but sustain extensive paralysis (approx. half of these occur above the sixth vertabra so affecting all four limbs)
most of these patients indicate that regaining the ability to grasp objects would provide the greatest practical benefit compared to regaining other lost functions
potential treatment for spinal cord injury
brain-machine interface
monkeys implanted with multi-electrode recording array in the hand area of M1
Separate surgery to implant intramuscular electrodes for recording and stimulation of hand and forearm muscles
Normal trials where data acquisition can start to predict the movements signalled by brain
Used neural predictions to control electrical stimulation of muscles having disabled the normal control of these muscles from brain
so could machines effectively replace outputs from the brain from motor cortex or even processing within the brain so frontal lobe?
functions of the frontal lobe
decision making attention impulsivity emotions old memories personality
aims of perception practical
does body physique relate to attributions to faces
Xu Lei
masculinity
attractivenessdoes context affect attraction to face cues
what stats
what methods for perceptual tests
defining how face shape changes with weight (BMI) - how were images made
took averages of 10 high BMI men and 10 low BMI men
what is BMI / problem
body mass index = fat and muscle
fat and muscle will affect face shape differently
fat and muscle predictions for masculinity and hypotheses
increased weight makes men look stronger
men have less fat and more muscle than women
hypothesis
increased muscle and fat increases masculinity
high muscle face shape looks more masculine than high fat face shape
muscle will enhance masculinity than fat
method for perception practical
transform face shape related to muscle and fat to 5 levels of muscle and 5 of fat
present faces
get ratings with a visual analogue scale low to high masculinity
what stats were used in the masculinity perception practical
calculated average ratings for each participant for 5 levels of fat and musce
tested if +4 transformed more masculine than low -4
carried out Wilcoxon t test
what were the results of perception masculinity practical
increasing fat and muscle increase masculinity
muscle increases masculinity more than fat
attraction perception practical hypotheses
choose more muscular (masculine) face for short than long term relationship
attraction perception practical method
face images - transform fat and muscle
choose most attractive transform level for long term relationship, short term sexual relationship
what stats carried out and result for attraction perception practical
paired t test
higher cues to muscle preferred for short compared to long term relationships
but higher level of fat cues not preferred for short compared to long term relationships
summary of three types of method for perception practical and positives / weaknesses
rating
-used for masculinity
-low sensitivity
-shows relative importance of different sues
method of adjustment or interactive choice
-used for attraction
-high sensitivity
-appropriate if optimal level of cue not know
forced choice between 2 alternatives (wasn’t used)
-high sensitivity
-need to know what to compare
what does r^2 tell us
coefficient of determination
gives a percentage
how much of one factor can be attributed to the other factor
