Week 1-5 content Flashcards
Empiricism
Knowledge is what can be observed / or ‘you gotta see it to believe it’
Operationalise
To formalise and define a measurement
Latent construct
Phenomenon that is not directly observed but are rather estimated from an operationalisation
Atomisation
Evaluate parts of the whole (in chemistry and physics, literally sub-atomic behaviour)
An operationalisation of a phenomenon should be:
Specific
(such as not measuring happiness by measuring how energetic a person feels)
Tied to specific theories/models of behaviour
(such as theoretical distinction between anxiety, fear and stress)
Developed in a replicable and recognisable style
(so that anyone could identify how someone arrived at this measure)
Falsifiability
The ability of an idea to be able to be demonstrably wrong
Hypothesis
A prediction about the nature of phenomenon
Null Hypothesis
There is no notable effect in the investigation.
Any small effects are due to chance alone.
Alternative Hypothesis
There is an effect of the investigation.
These differences are more notable than chance alone.
Inferential statistics/testing
Statistics used to make assessments of probabilities of ‘true’ effects (rather than just descriptions) of data
Research Questions
What is of interest in the current piece of work
HARKing
Claiming to have made a prediction after having seen the results of the test data
Variables
A variable is anything that varies between people, over time or within context: Number of apples eaten per week Ratings of happiness Alcohol tolerance Personality (e.g. extraversion) Average running speed Beliefs
Even labels or categories (i.e. nominal data)
Which football team you support or how you vote
Even how much you enjoy research methods and statistics will vary (across time and individuals!)
Variable Changes
Variables can change across time
Variables can differ between people
Variables can differ over contexts
What are not variables
Things that are NOT variables:
People (only the things that they do/think/feel)
Conditions or groups – but these might represent levels (i.e values) of variables
Random Variation
Sometimes called “Noise” or Error
Means that individuals differ naturally
Statistical tests try to account for or explain as much variation in data as it can
But there will always be some random Error that remains unexplained
Normal Distribution
X axis represents the scores on the scale
Y axis is the number of people who scored in each 5 point range (e.g. 51 to 55 = 19)
Natural variation
Some low scorers, some high scorers
Most people fall on the middle of the scale
If you were to randomly pick someone from the population what sort of score would they most likely have?
Independent Variables
Commonly noted as just “IV”
“Independent” because it’s value is fixed or controlled, and is NOT dependent on something else in our study.
A hypothesised cause – something that we think will have an effect on another variable
In an experiment, we manipulate the IV:
By assigning people to different groups and comparing them
By comparing people across different conditions
Levels of the IV
Levels are values taken on by an IV, usually categories of some kind
Usually, levels are simply the conditions or groups used in a study
Dependent Variables
Commonly noted as just “DV”
“Dependent” because it’s value depends on something else in our study (hopefully the IV)
A DV is a hypothesised effect – something that we think is affected by another variable
We always measure the DV
Nominal Data
Numbers used to represent names or categories or simple identifiers
Numbers used as categories or labels
The numbers have no mathematical value
Most statistical procedures don’t apply to nominal data (mean, standard deviation, etc.)
Some more advanced stats look at category membership
Ordinal Data
Order of performance.
The “distance” between each data point is not taken into account.
Ordinal data is used a lot in response scales to psychometric questionnaires (but is debated)
For example, a likert scale question on it’s own is ordinal, but when a questionnaire uses several questions to calculate a total score, this can be thought of as interval data.
Interval Data
Measurements with an equal interval between each point on the scale
Numbers on the scale can be positive or negative
Ratio Data
Ratio data also has measurements with an equal interval between each point on the scale
Also has a mathematically meaningful absolute zero
The absolute zero allows ratio judgements to be made (multiplication/division - e.g., Susan is twice as tall as her daughter)
Notice the similarity with Interval data
Operationalisation is also key – money in the bank v money in the pocket
Continuous and Discrete Variables
Ratio and Interval data are usually continuous
A variable is continuous if it can be subdivided further
10.62 cm makes sense
10.62th place in a marathon does not make sense
The opposite of continuous is discrete (not “discreet”)
Discrete data can be counted in whole numbers
Subdivision of the values doesn’t make sense
Number of siblings you have
Number of students in a class
Number of correct answers on an exam
A crisis of confidence
Prestige bias in publication
Prestige bias in resource concentration
‘Publish or Perish’ systems
Questionable Research Practices (QRPs; both systemic and personal)
Lack of transparency / data openness (‘file draw’ problems)
Internal Reliability
Within a study, how often do measures agree?
External Reliability / Replicability
Across a study, how often do findings agree?
Conceptual Replication – same idea, different method, different sample/data
Direct Replication – same idea, same methods, different sample/data
[Reanalysis of existing data is Reproducibility – same idea, same method, same data
Replication
Demonstrating the same findings as another study
Meta-Analysis
Statistical analysis of literature consistency
Systematic Review
A formal approach to evaluating arguments, practices and findings in the literature
Validity
The accuracy or realism of an operationalisation
Convergent Validity
Statistical agreement between two similar measures
Concurrent Validity
Statistical agreement between two measures of the same latent phenomenon
Discriminant Validity
The statistical ability to separate out target categories
Face Validity
Does this have the appearance of good measurement?
Content Validity
The value of the included/excluded content for representing a construct.
Generalisability
How representative a sample size or method is
W.E.I.R.D.
Western, Educated, Industrialised, Rich and Democratic Countries
Practicality/Utility
How useful a study is for an everyday application