PSYC210

Correlation

The relationship between two or more dependent variables. Does not include independent variables. Correlation can be used to describe and predict behaviour but not to explain or infer causality.

Positive correlation

As one variable gets bigger, the other variable gets bigger.
Graph tilts upwards from left to right.

Negative correlation

As one variable gets bigger, the other variable gets smaller.
Graph tilts downward from left to right.

Zero correlation

No consistent relationship between variables. Scattered points are in no pattern.

Correlations vary in terms of strength

The stronger the correlation the better the predictability.

Curvilinear relationship

An increase in x results has an initial increase in y, then a decrease in y., e.g., yerkes-dodson arousal curve.

Pearson's r

A way of expressing correlation. Ranges from -1 to +1.

Things to note about Pearson's r

- Variables to be correlated must be measured on the same individuals.
- Variables must be measured on an interval or ratio scale.
- It's linear because points generally fall on a straight line.
- For every change in x there is an equal constant change in y.

If r \= 0 (or low r)

There may be no relationship or it may be the existing relationship is non-linear.

Directionality problem

It is hard to determine if one thing determines the other or the other way round.

Cross-lagged-panel correlation procedure

A way of dealing with the directionality problem to a certain extent. Measure the correlation of two variables with each other over time. Can also look at the diagonals. If one variable causes the other it should be more strongly related over time.

Inferential statistics

Used to decide about the population based on observations of the sample.

Parameters

Characteristics of a population we are interested in. These are the mean and the standard deviation.

Three steps to sampling distributions and logic

- Make a guess about the population frequency distribution (mew - population mean).
- Take a random sample.
- Decide if the sample came from a population like the one you guessed in step one (usually based on how close the mean of the sample is to the hypothesised mean).

Hypothesis testing

Used to determine whether a population differs from a sample because of variability or a true difference.

Null hypothesis

The treatment has no effect and the difference we observe is due to variability and not a true effect.

Alternative hypothesis

The treatment has an effect and the difference is due to a true effect.

Significance level

How much different the samples must be from the population mean.

Region of rejection

The area of the distribution where you would reject the null hypothesis if the test results fall onto that area.

Critical values

Values on a distribution used to decide whether or not a statistical hypothesis test is significant or not.

Critical values and region of rejection

area on normal distribution, if the observed statistic is in region of rejection we reject null hypothesis and accept alternate hypothesis

Region of rejection/critical value on one-sided t-test

one critical value/area on graph for rejecting null hypothesis

Region of rejection/critical value on two-sided t-test

two critical values/areas on graph for rejecting null hypothesis

Increasing alpha level eg to 0.1

reduces type 2 error (false negative), more accurate, increases power, increases probability of correctly rejecting null hypothesis when it is false

Other impacts of increasing alpha (type 1 error)

increases type 1 error (false positive), set alpha very low eg 0.01 when consequences of type 1 error are severe eg drug trials in pregnant women, set alpha higher eg 0.05 when consequences are not as severe

Significance level or alpha level

The probability value that defines the boundary between rejecting or retaining the null hypothesis.

P < alpha

Reject the null hypothesis.

P \> alpha

Retain the null hypothesis.

One-tailed test

Used when there is evidence or theory to suggest that the treatment will have an effect in one particular direction.

Two-tailed test

- Used when there is another directional alternative.
- Used if there is no reason to predict the direction of effect.
- Alpha is still 0.05 but divided by two so the scores need to be higher or lower to hit the region of rejection.
- More conservative and less powerful.

Alpha

Probability the null hypothesis is true given the data.

Type 1 error

Rejecting the null hypothesis when it is true (alpha, false positive, eg conclude treatment is effective when it isn’t

Type 2 error

Retaining the null hypothesis when it is false (beta), false negative, eg conclude treatment is not effective when it is

How to reduce beta/type 2 error and increase power (1-beta)

- Increase alpha (reduces type 2 error and increases power).
- Increase n (less variability).
- Use most powerful statistical test.
- Have a good experimental design.

Estimated standard error of the mean

A useful number but we typically don't know the standard deviation.

Degrees of freedom

How many scores in the sample are free to vary- generally all scores except the last one.

Assumptions of the single-sample t-test

- The random sample comprises interval or ration scores.
- The distribution of individual scores is normal.
- The standard error of the mean is estimated using the s computed from the sample.

Steps for t-test

- Calculate the observed t using the estimated standard error of the mean.
- Determine the df.
- Look up the critical t in the t- table with the appropriate df (and alpha).
- If the observed t is greater than or equal to the tabled value than reject the null.

Two-sample t-test

Compare the difference between two sample means. Taking into account the sampling error by calculating the estimated standard error of the difference between the sample means.

Four goals of statistical analysis

- Describe: patterns of data.
- Explain: from hypothesis testing to real world.
- Predict: from models to future real world.
- Control: techniques to control for randomness.

Qualitative research

People and context.
- Interviews, observations.
- Themes.
- Results described in language.

Quantitative research

People and numbers.
- Patterns of numbers.
- Results described by statistics.
- Statistics translated into language.

Confirmatory research

- Describe
- Explain
- PREDICT
- CONTROL
Deductive, top-down.
Theory -\> hypothesis -\> pattern -\> observation.

Exploratory research

- DESCRIBE
- EXPLAIN
- Predict
- Control
Inductive, bottom-up.
Observation -\> pattern -\> tentative hypothesis -\> possible theory.

Confirmatory analysis

Prediction.
Control.
Description.

Central Limit Theorum

Different samples from the same population have different means.

Outliers

Individual data points that differ a lot from most of the others.

Conformity effect can be controlled by:

The independent variable.

Random variation and sampling error...

Cannot be controlled.

Controlled experimental study

- Experimenter directly controls changes in IV to observe how a DV changes.
- Exploratory.
- Confirmatory.
- Manipulation may have caused effect.
- Control for possible confounding factors or variables.

Correlational study

- No direct control.
- Relies on associations between variables that already exist.
- Often used in exploratory.
- Once aware of a relationship a researcher can design experiments that discover which of the two variables is causing the change in the other.

Multiple regression

Using more than one predictor variable.
Should improve predictive strength of our model.
Will also probably improve how well the regression equation model describes our sample data.

Neurons

Communication tools from one part of the brain to the other. Densely connected and have many dendrites.

Axons

Conduct electrical signals.

Myelin

A fatty substance surrounding axons. A major factor in determining the MR (magnetic resonance) signal and contrast.

The brain is organised into two types of 'tissues'

- Grey matter (composed cell bodies). Round the outside.
- White matter.

MRI is:

- Noisy. Measurements won't be perfect.
-Variable/configurable.

Analysis

- Based on statistics.
- Has many options/alternatives.
- Has more than one 'right' way (but many wrong).

Diffusion

Movement of particles across the space available to them.

Corpus Callosum

Connection between two hemispheres. Predominantly made up of white matter.

Gradient coils

Create magnetic field changes in any direction.

What does hydrogen do in an
MRI

Is present in the water. Hydrogen has magnetic properties and so has a magnetic moment where the ions line up in a specific direction.

Structural MRI

Images of gross brain anatomy at resolution

Structural MRI limitations

- Does not measure tissue type directly.
- The absolute values are not the same across all scanners.
- Measurement is in millimetres.
- Does not always distinguish bone from air.
- Contrast can be poor.
- A single sequence does not show all pathologies.
- Lots of artefacts and noise.

Complementary techniques to structural MRI

CT (with/without contrast).
- Shows bones, membranes, vessels, and tumours.
- Can't measure brain function.
Histology.
- Shows microstructure.

Structural vs diffusion MRI

Colours are different and distortion at front of the brain.
Structural MRI takes about five minutes to acquire one image.
Diffusion MRI takes lots of fast images (one every 1-3 seconds) for about five minutes.
Diffusion MRI has lower resolution (1-3mm).

Diffusion MRI

- Measures the direction of white matter fibres in the brain.
- Can give indications about integrity of the white matter.
- Can provide information on the connections between anatomical structures.
- Need to acquire many 'directions'.

Complementary techniques to diffusion MRI

Tracer studies (individual fibres in post-mortem brains).
Histology (myelin/axon dimensions/glia). Post-mortem only.

Functional MRI

- Measures brain activity by detecting changes associated with blood flow.
- Used to obtain functional information by visualizing cortical activity.
- Distortion in the front of the brain due to quick images.
- Can be undertaken when someone is performing a task or while someone is resting.

Structural vs diffusion vs fMRI

Structural MRI takes about five minutes to acquire one image.
Diffusion and fMRI take lots of fast images which are more susceptible to artifacts.
Diffusion and fMRI have lower spatial resolution than structural MRI.

The Haemodynamic response

The blood response to underlying neuronal activation.

Activated state

Oversupply of oxygenated blood to an area.
Increased cerebral blood flow and increase in cerebral blood volume, increase in oxygen supply, increase in MRI signal.

Experimenting with fMRI

Three types of events namely word generation, word shadowing, and null event. The null event acts as the baseline to compare with stimulation. The null event contains a visual component but no word component. Different tasks can be done and repeated to see what part of the brain lights up.

Verb fMRI experiment

If given a noun had to generate a verb, if given a verb had to shadow the verb.
Quite different areas of activation for the two tasks.
Generation activation \> shadowing activation.
Due to activation of Broca's area.

Broca's area

Contributes to speech production and fluency.

Limitations to functional MRI

- Does not measure electrical activity.
- Does not measure metabolic activity.
- Does not measure changes in blood oxygen levels resulting from electrical and metabolic activity of active neurons.
- BOLD (Blood oxygenation level-dependent) -fMRI is qualitative (change from baseline is important, not baseline itself).
- Sensitive to fast imaging artefacts.

Complementary techniques to fMRI

- Positron Emission Tomography (PET): measures brain metabolism directly. Can use radioactive tracers. Slower temporal resolution than fMRI. Reduced spatial resolution than fMRI.
- Electroencephalography (EEG): measures electrical brain activity directly. Much faster temporal resolution than fMRI (milliseconds). Reduced spatial resolution than fMRI.

Qualitative data

- Quotes or images.
- Capturing the original quality of the data.
- Non-numeric.
- Primarily involves the human experience.

Theme

A patterned response or meaning within the data set.
- Category.
- Dominant discourse.
Not simply a topic discussed by the participant.

Ontology

Views on human reality

Quantitative research (realism)

There is one 'true' reality (independent of perception)

Qualitative research (relativism)

People's realities differ (relative to perception).
Our knowledge of reality is never a simple reflection of the way the world actually is, but is created and sustained through subjective social processes.

Epistemology

What we know and how we know it

Quantitative research (positivism)

Knowledge and meaning is waiting to be discovered and is then considered 'true' until disproven (through research).

Qualitative research (social constructionism)

Knowledge and meaning is being generated by attempts to explain the human world (including research)

Epistemology key points

Researchers should know which epistemology they are using.
Different approaches to research generate different knowledge.
There are more than these two.

Why use qualitative research?

If it involves people then the meaning matters.
Numbers can't always do this.

Five key elements of qualitative research questions

- State a goal.
- Define the population sample.
- Define the setting.
- Identify the primary topic.
- Be precise enough to be feasible.

Closed questions

Imply fixed answer choices.

Open-ended questions

Invite expansion.

Three types of interview structure

- Structured (quantitative)
- Unstructured (research development and pilot studies)
- Semi-structured (qualitative).

Structured interviews

- Closed questions.
- Very fixed topic and fixed order of questions.
- Very clear roles.
- Expansion allowed only if pre-defined \= branching.
Outcome: specific answers + numerical data \= quantitative.

Semi-structured interviews

- Open-ended questions (or probing following closed questions).
- Very open around a topic and question order can vary.
- Almost equal roles.
- Expansion encouraged (on topic).
Outcome: open-ended answers + verbal data \= qualitative.

One-on-one interviewing

- Most common source of qualitative data.
- Smaller samples are better.
- Smallest possible sample is a case study (n\=1).
- Case studies allow exploration of the one case in depth.
- Most one-on-one interview studies have more than one participant.

De Visser and Smith (2006) aim

Discourse: patterns of language about a concept.
Identity: notions of self that draw upon available discourses.

De Visser and Smith (2006) three key issues

- Relationships between different discourses of masculinity.
- Health-related behaviour in the performance of masculine identities.
- Meanings of masculine behaviour and masculine identity to young men.

De Visser and Smith (2006) key concept

Hegemonic masculinity: expected patterns of behaviour among men.
- Playing or watching sport.
- Drinking alcohol.
- Predatory masculinity.
- Interpersonal violence.

De Visser and Smith (2006) research questions

How do men come to identify with a particular discourse of masculinity?
What does it feel like for a man to reject hegemonic masculinity and manage an alternative masculine identity?

De Visser and Smith (2006) participants

Sample \= one participant (case study).
Rahul: first year undergrad (19) studying in London.
Born in the UK (parents in India).
Not rich but went to a private school.

De Visser and Smith (2006) method

Semi-structured interview about how Rahul spends his free time. Prompts about:
- Drinking alcohol.
- Sexual activity.
- Exercise and sport.