f

Standard deviation: shows the amount of variation in a data set, assessing the spread of data around the mean.

Tells us the quality of the mean in terms of how well it represents the rest of the scores.

A large SD tells us that the scores are widely spread out above and below the means suggesting is not very representative of the rest of the scores. A small SD means the mean is representative of the scores from which it was calculated.

+ More precise/sensitive as all values are taken into account and is not heavily distorted by extreme scores.

- However this may hide some of the characteristics of the data set (e.g. extreme values).

- Complicated to calculate.

- Less meaningful if data not normally distributed.

The Range: The difference between the highest and lowest number (highest no. – smallest no.)

+ Easy to calculate.

+ Provides you with direct information.

- Affected by extreme values.

- Doesn’t take into account the number of observations in the data set.

- Tells us very little about the actual spread of scores e.g. how spread out or clustered they are.

Semi interquartile Range: The measure of the spread of middle 50% of scores avoiding the extreme scores that may be in top and bottom 25%. Usually used when median is the measure o central tendency because of the similarities in their calculations as they are by the central scores.

+ Less sensitive/distorted to extreme scores than the range.

- Only uses half the data so much of the data doesn’t add anything to its calculation.

- Laborious to calculate by hand as it involves ranking etc.

Which measure of dispersion to use with a measure of central tendency?

A measure of central tendency should always be accompanied by at least one measure of dispersion. The choice of which to use is down to careful consideration of the raw data that has been gathered however the table below is a rule of thumb.

When using a ….use a ….

Mean - Standard deviation

Median - Semi-Interquartile range/Range

Mode - Range

Tables: Display a clear summary of raw data (numbers which haven’t been treated in anyway)

A simple way and clear to present data is to put them in tables. How you construct a table will depend on the kind of data gathered and the R method used. It’s usual to use measures of central tendency and dispersion in a table rather than raw numbers. Which you chose depends on a consideration of both the kind of data you have collected and the advantages and disadvantages of each option. A title explaining what the numbers are and what they represent is essential if the table is to effectively communicate these findings.

Graph: A method used to convey information pictorially and clearly and therefore care should be taken in choosing and drawing the graph e.g. label and title the graph and avoid putting looking good above being simple and clear.

-Data shouldn’t be presented in ways that are misleading e.g. the distances between points on a vertical axis should be equal and the scale not exaggerated and distorted by the look of the graph. Since graphs are not intended to summarise data it’s not appropriate to use individual; P scores useless you are constructing a scattergraph. If data is gathered using the experimental method, usual convention is to plot the DV on the vertical axis and the IV on the horizontal axis.

Line graphs:  Display numerical data but not categorised data. As with the bar chart, the y axis represents frequency but the values along the x must be continuous.

Bar charts: display data in categories with gaps in between the categories. The height of the bar represents the frequency.

• Used for nominal data (data in categories) discreet data- when data fits into 1 category only e.g. lion.
• The x-axis (frequencies are usually on the y-axis) does not need to show a complete scale (if showing categories)
• There should be gaps between the bars.
• When drawing a bar chart, the vertical axis should show the score of the variable e.g. the mean/frequency whilst the horizontal axis should show the individual categories/variables you measured. The bars on the horizontal axis should be drawn separately with equal width and gaps. Need not show all the categories on the horizontal axis, it’s acceptable to just show those of interests as a comparison. However being selective in this way can be misleading so are must be taken as only choosing to show certain categories doesn’t tell the whole story

Histograms: use continuous data and so there are no gaps between the vertical bars; this indicates that the horizontal axis has a continuous measure rather than distinct categories. The vertical axis represent the frequency of something has occurred. The points on the vertical axis should be equal as should the width of the columns.

• Used for interval or ordinal data.
• No intervals (if data is grouped) are missed, even if they are empty. Class intervals are represented by their mid-point at the centre of each column. There are no gaps between columns.

Frequency polygons: Like a histogram but the midpoints joined by a straight continuous line, highlighting continuous nature of the variable on the x axis. 

• Used for interval or ordinal data.
• All class intervals are represented.
• Instead of columns, a line is used to join the mid-point of each class interval.

Non-numerical - feelings. Cant be quantified but can turn into quantative data by use of categories.

+ Rich/detailed data of emotions, opinion  that may not be assessed using quantative methods with closed q's.

+ Represents the true complexities of human behaviour

+ Useful for studies at individual level, to find in depth, the ways in which people think /feel (e.g. case studies).

+ Takes the point of view of P as their responses are not restricted in advance by the point of view of the R.

+ Provides rich details of how people behave because P's given free range to express themselves

- Is less controlled and structured compared to quantative so is hard to assess reliability.

- Diff and laborious to analyse and difficult to see patterns in the data that would allow you to draw conclusion.

- Subjective analysis can be affected by personal expectations/beliefs (though quantative methods may only appear to be objetive but are equally affected by bias.

R techniques produce this data: (unstructured) observations, content analysis, case studies, questionnaire/interviews (with open q's) e.g. unstructured interview

Is challenging to summarise as there is lots of it i.e. video recordings and large amount of written material. But must find ways of summarising data to draw conclusions

The first step is to categorise data is some way:

Pre-existing Categories: i.e. the R decides on the categories before beginning the research

Emergent Categories: i.e. the categories/themes emerge when examining the data

Later the behavioural categories can be used to summarise the data

• The categories/themes could be listed
• Examples of behaviours within the category may be represented using quotes from P’s or descriptions of typical behaviours in that category
• Frequency of occurrences in each category can be counted and turned into quantitative data
• Finally a researcher can draw conclusions

Multi-method approach:Combination of different techniques and methods.

Meta-analysis:R studies findings from number of different studies in order to reach a general conclusion about a particular hypothesis. The R uses effect size (measure of strength of relationship between 2 variables) as DV.

+ More reliable conclusions can be drawn.

- Research designs vary, so studies can never be truly comparable.                

Cross-cultural studies:Natural experiment in which the IV is different cultural practices and the DV is a behaviour e.g. attachment.

+ Does allow R’s to see if some behaviours are universal.

- Imposed etic (tests developed in one country are used in another, with different norms, so affect results). Also group of P’s may not be representative of whole culture, but generalisations are made.

Longitudinal studies: Observation of same items over long period of time, usually aiming to compare same individuals at different ages, to observe long term effects.

+ High in validity, people usually do not remember past events and if they were asked about their past, they would not remember.

- Attrition: (loss of P’s over time, leaving biased, small sample) is a problem.

- Also P’s are likely to become aware of aims.

- Subject to cohort effects (one P may have unique characteristic due to specific experiences) so the group studies may not be typical.

Cross-sectional studies:One group of P’s of young age are compared with another, older group, with the view of finding out the influence of age on the behaviour.

+ Quick and easy and relatively cheap way to gather data.

- P variables not controlled and cohort effects can mean the groups are different and so are difficult to compare to each other.

Role play: Controlled observation in which p’s are asked to imagine how they would behave in certain situations.

+ Enables R’s to study behaviours which might otherwise be unethical.

- It may not be an accurate and valid representation of how people would act.

• Internal reliability:  measure of the extent to which something is consistent within itself. For example, all the q’s on an IQ test (which is a kind of questionnaire) should be measuring the same thing. May not be relevant to all questionnaires because sometimes internal consistency is not important e.g. a questionnaire about day-care experiences may look at many different aspects of day care and is effects.
• External Reliability: measure of consistency over several different occasions. For example, if an interviewer conducted an interview, and then conducted the same interview with the same interviewee a week later, the outcome should be the same
• Reliability also is whether 2 interviewers produce same outcome - inter-interviewer reliability .
• Assessing Reliability - >
• Internal reliability - Split half reliability: Splitting test into 2 halves, carry out a correlation on the 2 halves, to ensure both halves of test are of equal difficulty, if they are, test is reliable. E.g. single group of P's all take test at once their answers are split in half and this could be done by comparing answers to odd number q’s with all the answers to the even number q’s. The individual scores on both halves of the test should be very similar and the 2 scores can be compared by calculating a correlation coefficient.
• External reliability -Test re-test reliability: Experiment administered twice, with a gap in-between. If the same P’s score similar results on both occasions, method is reliable. Used for factors that are stable over time, e.g. intelligence. Reliability higher when little time passed between tests. If test produces scores these compared by calculating correlation coefficient.

• Validity: extent to which something measures what it is supposed to measure. Involves issues of control, realism and generalisability. But study can have high realism but lack generalisability.
• Control: The extent to which any variable is held constant/regulated by a R it is important to control as many relevant EV’s as possible otherwise results would be meaningless as the R may have not actually tested what they intended to and instead the influence of another variable not the IV has been tested. 
• Mundane Realism: Refers to how a study mirrors the real world. The simulated task environment is realistic to the degree to which experiences encountered in the environment will occur in the real world
• Generalisability: Just because a study is conducted in a natural environment it does not mean you can generalise the findings to the real world e.g. if only use US uni students ‘though experiment may be in a natural setting’ can’t generalise results to all ages/cultures.
• Internal validity (about control and realism) : extent to which study measures what it is set out to measure/degree to which the observed effect was due to the experimental manipulation rather than other factors such as EV'S which may affect results.
• External validity (about generalisability): The degree to which the findings can be generalised to other settings (ecological validity), to other groups of people (population validity) or over time/to any era (historical validity). Can be affected by representativeness of the sample.
• External Validity is affected by internal validity – cant generalise R' s of study if low in internal validity.
• Sample validity:The extent to which the P’s represent people outside the R situation.

• Aim:  Intended purpose of investigation
• Hypothesis: Precise, testable statement written in future tense about target pop.
• Operationalised Hypothesis: Make it testable, so it can be repeated, increasing reliability of findings. Must operationalise the V's (IV and DV–ensuring  in a form that can be easily tested) – e.g. how will you measure IV/DV. E.g. 'scores obtained on a memory test by 10 females aged 16 will be higher than the scores obtained by  10 males aged 16-24'.
• Research hypothesis: Proposed at the start of R and is often based on theory
• Experimental (experiments, H1)/Alternative Hypothesis (observations/opinions, HA): The prediction you are making e.g. evidence there is a significant relationship/ difference between two sets of data.
• Null H (Ho): Backup H; statement of no diffrelationship. If data not support Ho, reject it go with HA instead.
• Directional hypothesis (One-tailed Hypothesis): States expected direction of the predicted difference b/ween two conditions or groups of P’s  e.g. P’s do hmwk without music produce better results than P’s who do hmwk with music. Previous R/pilot study may suggest direction. Easier to reject than a non-directional, so R that proves a directional hypothesis is regarded highly.
• Non-directional hypothesis (Two-tailed Hypothesis): Predicts that there will be a difference but not the direction of the difference between two conditions or groups of participants ; e.g. P’s who do homework without music will produce different results to P’s who do homework with music. (Use if don’t know answer to prob or think something may happen – next piece R may choose directional.)