Research methods

what does partitioning deviations allow us to do

compare variability due to our treatment with variability due to other factors

what is the equation for within-groups sum of squares

the total deviation of each exam score from the group mean squared

what is the equation for between-groups sum of squares

the total square of how much each group mean deviates from the grand mean squared

total sum of squares

how much each exam score differs from the grand mean - sum of squares within groups added to sum of squares between groups

what does the degrees of freedom associated with sum of squares represent

the number of independent values that can vary in a statistical calculation after all constraints have been imposed on the data

what does n stand for in DOF calculations

number of participants

what does k stand for in DOF calculations

number of groups

degrees of freedom total

d=n-1

degrees of freedom within group

d=(n-k)

what do we do to calculate DOF when the groups have different numbers of participants

quantify degrees of freedom for each group and add them together

degrees of freedom between groups

df=k-1

how to calculate mean square for within, between, total

sum of squares / degrees of freedom

what is mean square the same as

variance

how to calculate f ratio

variance between groups / variance within groups

what does variance between groups / variance within groups also mean

(treatment effect + experimental error) / experimental error

what does the f ratio allow us to do

estimate the treatment effect after accounting for error

what about the f ratio allows us to reject the null hypothesis

being greater than 1

what do we compare the f ratio to

f-distribution

what does f-distribution depend on

numerator dfA

denominator dfR

what causes f-distribution to become more skewed

smaller degree of freedom

what happend when between groups df is fixed and within groups Df increases

distribution effect shifts to the right

what happend when within groups df is fixed and between groups Df increases

distribution effect shifts to the right

how can we reject the null hypothesis

if f-ratio exceeds the critical value

what do columns on the f-tables represent

effect / between-group degrees of freedom

what do rows on the f-tables represent

error / within-group degrees of freedom

what do different f-tables represent

different alpha levels

how do we calculate if a skewness value is significantly different from 0

Z-test

how is Z-test calculated

skewness / standard error of skewness

what is SEy

standard error associated with skewness

how is SEy calculated

square root of 6 / n

purpose of transforming data

transforming data to meet normality or homogeneity of variance

what must we do when transforming data

must apply the same transformation to all groups or conditions

requirements of transforming data

maintain the ranked order of data - monotonic transformation

must not violate other ANOVA assumptions

what is moderate skew

1.96< z < 2.33

what is appropriate transformation for moderate skew

positive = square root of y

negative - square root of k-y

what is substantial skew

2.33 < z < 2.56

what is appropriate transformation for substantial skew

positive - logy

negative - log(k-y)

what is severe skew

z>2.56

what is appropriate transformation for severe skew

positive - y^-1

negative - (k-y)^1

when to transform

When there is substantial skew / heterogenity of variance

Only transform when the violation is substantial

effect of transformation on type II error

Increases statistical power, increasing ability to detect real effect

Reduces chance of missing true effect

effect of trasnformation on type I error

ANOVA is robust against minor violations

Not transforming when you should won't increase false positive rate

epistemological standpoint

essentialist / realist

critical realist

essentialist / realist

focus on experiences, meanings, reality of lives

words provide direct access to a participant’s inner world

critical realist

how individuals make meaning of experience based on their their socio-cultural situation

words provide access to the participant’s version of reality

theme

Patterned response in data set

Has meaning in relation to research question

No rules to classify

Prevalence may or may not be important - depends on research question

types of thematic analysis

inductive

deductive

inductive thematic analysis

data-driven

aimed at extracting themes grounded in data, what the participant actually said

deductive thematic analysis

using existing theory to guide the analysis and extraction of themes

moves beyond the semantic meanings offered in the data set

whos framework guides thematic analysis

braun and clarke (2006)

steps of thematic analysis

familiarise self with the data

generate initial codes

search for themes

review themes

define and name themes

produce

what is coding

process of identifying aspects of data which relates to research question

semantic codes

Succinct summary of explicit content of data

Based in semantic meaning

Close to content of the data and participant's meanings

latent codes

Beyond explicit content of data, provide interpretation about data content

Invoke researcher's conceptual and theoretical frameworks to identify implicit meanings withing data

familiarise self with data

Get to know data

Repeated reading of whole data set

Active reading

End phase - make notes on overall observations on data set

generate initial codes

Inclusive, thorough, systematic - work through each data set item before proceeding to the next

Provides building blocks of analysis

Approach to coding depends on type of TA

End phase with data coded and codes collated

searching for themes

Organise different codes into potential themes

Start with list of collated codes

Review codes and collated data relating to each code

Cluster codes - share a unifying feature so they reflect and describe a coherent and meaningful pattern in the data

Start to explore relationship between themes and how themes will work together to tell overall story about the data

End phase with thematic map / table outlining candidate themes

Collate all data extracts relevant to each theme

Review themes

Do tentative themes form a coherent pattern?

Codes within a theme should fit together meaningfully, be relevant to research question

Each theme should be distinct from another one

Re-read whole data set - codes may get discarded, new data coded, theme boundaries may get reworked

define and name themes

Clearly define themes - be able to clearly state what is unique and specific about theme

Determining what aspect of the data each theme captures

produce report

Construct narrative that foes beyond description of the data

Make and argument for how themes relate to research question

How codes within a theme are related

Provide support for themes - limited but vivid data extracts to support your argument

deontological ethics

duty and rights-based

actions are considered right or wrong depending on whether they are consistent with the duties of the agents and the rights of those affected by the actions

example of deontological ethicalist

kant

consequentialist ethics

outcome-based ethics

actions are considered right or wrong following the weighing of positive and negative outcomes

example of consequential ethicist

bentham

guiding principles of research with human participants

autonomy and protection

history of ethics of human participants

Nuremberg Code - Nuremberg military tribunal 1947 - emphasis on informed consent

Declaration of Helsinki - World Medical Association 1964 - outlines ethical principles involving human ppts - mostly medical practice

primary ethical principles of the BPS code of ethics and conduct

respect for dignity, autonomy, and privacy

competence

responsibility

integrity

when might psychological research involve animals

fundamental behavioural / cognitive mechanisms

neurobiological mechanisms of behaviour

guiding principles of humane experimentation - Russel and Burch’s The Three Rs

Refinement - reduction in severity of inhumane procedures

Reduction - in number of animals used

Replacement - of highly sentient animals whenever possible

Law requires researchers to follow these principles and animal welfare

Animal Scientific Procedures Act (1986)

regulates all animal experiments involving vertebrates and octopi

When is animal research permitted

‘designated establishments’ under remit of ‘project licences’ by researchers that have completed accredited training programmes to obtain a ‘personal licence’ (controlled by the Home Office

How to handle data

All steps of data collection and treatment must be carefully documented

Data must be stored in such a way that they can be retrieved for later verification

What does authorship imply

Important contribution to plnning, execution, evaluation of research

Contribution to manuscript and the approval of the final version

Conflict of interest

Any situation in which financial / personal considerations have the potential to compromise scientific or professional conduct

how can power analysis help us plan study requirements

sample size

resource allocation

balancing feasability with scientific rigour

how can power analysis help us evaluate study sensitivity

assess its ability to detect meaningful effects

determine confidence in non-significant results

support interpretation of results

what do we need to know before starting a study

the size of the treatment effect we are looking for (expected magnitude, clinical significance)

how certain we need to be about our findings (alpha, power)

how many participants we need (sample size requirements for reliable results, practical constraints)

effect size

tells us the magnitude and practical significance of research findings

what does increased population difference mean with the same variance for effect size

bigger effect size

what does increased population variation with the same mean difference mean for effect size

bigger effect size

measures of effect size

eta-squared

partial eta-squared

r-squared

what does eta-squared mean

proportion of variance in dependent variable explained by a single IV

how is eta-squared calculated

sum of squares between groups / sum of squares total

partial eta-squared

estimates a specific effect size if there is more than one independent variable

how is partial eta-squared calculated

sum of squares between / (sum of squares between + sum of squares within)

r-squared

proportion of variance explained by the model

how is r-squared calculated

1-(sum of squares within/sum of squares total)

measure of difference

cohen’s d

cohen’s d

when there are only two groups, a ratio of the difference between the two groups with their error variance

how to calculate cohen’s d

(mean 1-mean 2) / square root of mean squared error

benchmarks for eta squared (and partial)

small = 0.01

medium = 0.06

large = 0.14

benchmarks for R-squared

small = 0.01

medium = 0.09

large = 0.25

benchmarks for cohen’s d

small - 0.2-0.3

medium - 0.3-0.5

large - 0.5+

 two ways to decide what effect size is being aimed for in the planning stage

Based on previous research - meta-analysis - review previous literature, calculate previously observed effect sizes from the same / similar studies

Based on theoretical importance - decide whether a small / medium / large effect is required

how effect size is used in the evaluation stage for interpretation of the results

Supports clinical / practical decision making

Facilitates meta-analysis and replication

equation for power

1-beta

beta

probability of making a type 2 error (false negative)

factors determining the power of a study

alpha level - more conservative, lower power

effect size - larger, higher power

sample size - larger, higher power

what do we need to do if we have a smaller effect size

larger samples to acheive the same power

what do we need to do to estimate required sample size

hold other factors constant