Methods glossary Flashcards
1
Q
<p>Alpha value</p>
A
<p>Usually, the probability value
of .05 is the alpha value used in inferential statistics as the measure of significance. This value has been adopted as the threshold for accepting the null hypothesis (test result is greater than .05), or rejecting it (test result is equal to or less than .05).</p>
2
Q
<p>Bar graph</p>
A
<p>A graph often used to visually compare differences between the means of separate groups or conditions. The x-axis represents the different groups and the y-axis represents the means. Each mean is drawn as a vertical bar, and because the means are from different categories (groups or conditions) the bars are not joined.</p>
3
Q
<p>Baseline measure</p>
A
<p>The effect of the control condition on the dependent variable against which the effect of the experimental condition can be compared to give the
size of the difference between the two conditions.</p>
4
Q
<p>Behavioural data</p>
A
<p>Data produced from measuring behaviour. These can cover a wide range of activities such as reaction times (e.g. time taken to press a buzzer) and memory (e.g. number of words correctly recognised) as well as less well-defined behaviours such as problem solving which might be described qualitatively.</p>
5
Q
<p>Between-participants design</p>
A
<p>An experimental design where different participants complete each condition, so more participants are needed than for a within-participants design. Also known as independent groups design, independent
samples design, or independent measures design. Can help eliminate confounding variables such as demand characteristics since participants’ understanding of the whole experiment is restricted, but does not reduce the influence of individual differences.
</p>
6
Q
<p>Bimodal distribution</p>
A
<p>A distribution with two modes. The distribution is described by two symmetrical bell-shaped curves that appear joined, with two peaks representing two values for the mode.</p>
7
Q
<p>Categorical data</p>
A
<p>Data that have been classified into discrete categories which are measured at the nominal level. Numbers are often used as labels (e.g. male = 1, female = 2), but the order numerically is of no importance.</p>
8
Q
<p>
| Categorical variable</p>
A
<p>A variable which is measured at the nominal level; data produced are grouped into mutually exclusive and distinct categories.</p>
9
Q
<p>Cause and effect</p>
A
<p>The aim of any experiment, a general law which can be established when an isolated, independent variable is manipulated to cause a measurable effect on the dependent variable.</p>
10
Q
<p>Chi-square test</p>
A
<p>A statistical test used to analyse data measured at nominal level. This test allows one to look for associations between two categorical variables, by comparing the observed frequencies against
the expected frequencies (see contingency 2
table). The test calculates the statistic c and Cramer’s V provides a measure of effect size.</p>
11
Q
<p>Coding</p>
A
<p>The process of assigning or converting material to a code for the purpose of identification, classification or analysis.</p>
12
Q
<p>Complete observer</p>
A
<p>A researcher who openly observes but does not participate in the research setting.</p>
13
Q
<p>Condition</p>
A
<p>In an experiment, the different forms of the independent variable created from its being manipulated. Very often an experimental condition and a control condition are set up, so that the effects of each can be measured and compared, with the control condition giving a baseline measure.</p>
14
Q
<p>Conditional Probability</p>
A
<p>The likelihood of something happening that is dependent on something else.</p>
15
Q
<p>Confidence interval</p>
A
<p>The range of values within which a population mean is likely to fall. Confidence intervals are specified by stating the lower and upper bounds of the range. For normally distributed data, we can be 95 per cent certain that the population mean will fall within 1.96 standard deviation points of a sample mean. This allows us to judge whether two samples are from the same population where any difference between them is simply due to sampling error.</p>
16
Q
<p>Confounding variable</p>
A
<p>A variable, which is not the independent variable, that affects the dependent variable in one condition more than another – hence confounding the results. Researchers strive to eliminate confounding variables through good experimental design.</p>
17
Q
<p>Content analysis</p>
A
<p>A quantitative method of analysing data. For example, data from an interview will be analysed by counting the prevalence and sequence of certain words
and these are sometimes analysed using a chi-square test.</p>
18
Q
<p>Contingency table</p>
A
<p>Table used in studies looking for an association between independent (mutually exclusive) categorical variables. The table presents the number of observations in each possible combination, or contingency, of each category. If there are two variables (gender; film preference) each with two categories (male/female; horror/romance), then the number of observations in each possible combination of categories would be presented in a 2 x 2 contingency table.</p>
19
Q
<p>
| Continuous variable</p>
A
<p>A variable that can produce data of any value (including decimal places) between the highest and lowest points on a scale; e.g. time taken to recall items in a memory test.</p>
20
Q
<p>Control condition</p>
A
<p>The condition of the independent variable which is in all respects but one identical to the experimental condition – nothing is introduced to cause the dependent variable to change. The control condition can then be used to give a baseline measure.</p>
21
Q
<p>Controls</p>
A
<p>Techniques used in an experiment to eliminate the foreseeable effects of any confounding variables.</p>
22
Q
<p>Conversation analysis (CA)</p>
A
<p>An analytic method which focuses on the precise details of conversational interactions and on how people talk – for example the orderliness, structure and sequential patterns of interaction.</p>
23
Q
<p>Correlation</p>
A
<p>The relationship, or association, between two variables whereby if the value of one changes so does the value of the other. One variable cannot be said to cause the other to change.</p>
24
Q
Correlation coefficient
A
A numerical measure between -1 and +1 where the size of the number indicates the strength of the relationship between the two variables in a correlation, or the effect size. A value of 1 shows a perfect positive correlation while a value of -1 shows a perfect negative correlation. If the correlation coefficient is 0, there is no relationship between the two variables at all.
25
Counterbalancing
A control introduced to eliminate practise effects as a confounding variable in a within-participants experiment, whereby different participants undergo the conditions in a different order from one another.
26
Cramer’s V
A measure of effect size reported when using the chi-square test.
27
Data
Information collected during a study that has been documented in a suitable manner ready for communication or analysis.
28
Debriefing
The procedure where participants are given previously undisclosed information about a research project following completion of their participation.
29
Degrees of freedom (df)
A concept often defined as the number of scores whose values are free to vary. Degrees of freedom take account of the size of a sample, the number of variables, and the number of conditions in a study, and are reported with most inferential statistics.
30
Demand characteristics
Potentially a confounding variable. Introduced when participants do not respond naturally in an experiment, but react as they think they are expected to. The influence of demand characteristics may be reduced by using a
between-participants design, as participants take part in only one condition.
31
Dependent variable (DV)
The variable that is measured in an experiment, whose values result from manipulating the independent variable.
32
Descriptive statistics
A set of statistics used to summarise and present numerical data (e.g. mean and standard deviation).
33
Discourse analysis
Focuses on the ways in which people make meaning – discourse is unravelled in microscopic ways to see what it consists of and how it is put together to accomplish different actions.
34
Discrete variable
A variable that can produce data of only certain values, such as only whole numbers, between the highest and lowest points on a scale; e.g. the number of items recalled in a memory test can only be measured in whole numbers.
35
Double-blind testing
In drug trials particularly, the process of not informing either the participants or the drug administrator whether they are dealing with the real drug or not, to avoid confounding variables such as demand characteristics.
36
Ecological validity
The extent to which a study reflects naturally occurring or everyday situations. Ecological validity is often low in laboratory experiments where control of confounding variables is, however, higher.
37
Effect size
The magnitude of the effect the independent variable has on the dependent variable or the size of the relationship between two variables. In experimental designs, the statistic d is a standardised measure of effect size, calculated by subtracting one sample mean from the otherand dividing the result by the mean standard deviation. A correlation coefficient is also a measure of effect size, as is Cramer’s V which is a measure of effect size for a chi-square test.
38
Empirical evidence
Evidence based on some form of experience such as that gained through experimentation, observation, surveying or interviewing.
39
Epistemology
Epistemology is the ‘theory of knowledge’. It refers to the principles of what can be known and how we can know it; that is, how we can find out about it.
40
Error bar chart
A graph used to present confidence intervals. The upper and lower bounds of the confidence interval are represented by single lines above and below the mean, which is drawn as a single point.
41
Evidence
Information presented to determine or demonstrate the truth or the plausibility of an assertion, a consequence of the way that information is gathered, documented and analysed. Anecdotal evidence is informal argument based on personal experience or hearsay. In quantitative research, scientific evidence serves to either support or
counter the truth and validity of a hypothesis and is established through empirical analysis such as statistics. In hermeneutic research, evidence means drawing plausible inferences from patterns of meaning implied in the data.
42
Experiment
A research method used to investigate a hypothesis about the effect of the independent variable on the dependent variable. As far as possible the experimenter attempts to control for any other variables that might influence the dependent variable.
43
Experimental condition
A form of the independent variable created from its being manipulated. Unlike control conditions, something is introduced to the experimental condition as the possible cause of change in the dependent variable.
44
Experimental design
See ‘between- participants design’ and ‘within-participants design’.
45
Experimenter effects
Potentially, a confounding variable. Introduced by the experimenter or researcher exerting an influence, consciously or unconsciously, reducing objectivity. Experimenter effects can be overcome by ensuring all participants are treated in the same way as each other (e.g. each participant is read the same instructions), and by limiting experimenter contact with them as far as possible.
46
Fatigue effects
Introduced when a study takes a long time to run and participants become tired. Fatigue can affect responses as the study progresses, particularly if the participants are young children or elderly people. Fatigue effects can sometimes be limited in experiments by using a between- participants design, so no individual participant is required to complete the whole experiment.
47
First-order coding
The organising and categorising of data in a transcript by capturing chunks of meaning and giving them codes or labels. The process is purely descriptive, with minimal interpretation.
48
Fisher’s exact probability test
A statistical test used instead of the chi-square test when an expected frequency of less than 5 occurs in more than 25% of the cells of a 2 6 2 contingency table.
49
Generalisability
The extent to which research findings may be applied to a larger population and/or other situations.
50
Generalisation
To establish general laws or statements which apply across participants and specified conditions.
51
Habituation
Introduced when the same stimulus or task is presented to the same participants many times. Participants become increasingly used to responding to the stimulus or performing the task, which may obscure the effects of the independent variable. Habituation can sometimes be limited by using a between-participants design, or by simply reducing the number of responses required of participants.
52
Hermeneutic
The theory and practice of interpretation, involving the analysis of meanings in subjective accounts.
53
Histogram
A graph showing the number of occurrences, or frequency, of each value in a data set across the range (hence also known as a frequency distribution). The y-axis represents frequency and the x-axis represents the data range. Each value is drawn as a vertical bar and, because data are continuous, the bars are joined together.
54
Hypothesis
A statement making a prediction about the results of a study.
55
Independent samples t-test
A parametric test of difference used in inferential statistics to analyse data produced by an experiment with a between-participants design and which has two conditions. This test may also be called the unrelated t-test, independent t-test or between-groups t-test.
56
Independent variable (IV)
The variable that is manipulated in an experiment, to
| investigate a potential effect on the dependent variable that is measured.
57
Individual differences
Introduced when, for example, different participants are allocated to different conditions in an experiment, so that the naturally occurring variations between people may obscure the effects of the independent variable. The influence of individual differences may be reduced by firstly ensuring the sample is large enough, by randomly allocating participants to each condition, and/or by using a within- participants design.
58
Inferential statistics
These involve statistical tests which allow conclusions to be drawn from numerical data generated by research. The tests calculate a statistic, from which a probability value can be derived telling us the likelihood of any difference of relationship being due to sampling error.Informed consent
59
Informed consent
The permission granted by a participant after he or she has received comprehensive information about the study.
60
Inner experiences
The private monologues (thoughts), feelings, sensations and process experienced by individuals either at a conscious or unconscious level.
61
Insider viewpoint
A viewpoint gained from introspection, interviews and analyses, where the researcher tries to see and think about the data through the eyes of the participant generating the data.
62
Interpretation
A representation of the meaning or significance of something.
63
Interpretative repertoires
A term associated with discourse analysis. Interpretative repertoires are ‘turns of phrase’ or arguments that people typically employ to describe or explain things. These are systematically related sets of terms that are commonly used in society because they are developed from particular historical contexts and have become part of the common sense of a culture or a particular institution. They are often organised around metaphors. This means that interpretative repertoires are a way of picking out the familiar arguments that tell us something about how a society, community or social group make sense of social life. In other words, they allow us to identify the taken-for-granted cultural resources found in talk.
64
Interval level scale
A numerical scale where the intervals between measurements are specified and constant but the scale does not contain a true zero (e.g. Celsius temperature measurement scales), so the numbers used are arbitrary representations (e.g. 80 degrees Celsius is not twice as hot as 40 degrees).
65
Interview
A ‘conversation with a purpose’, interviews provide in-depth information or understanding about a particular research issue or question. Structured interviews consist of a list of specific quotations where the interviewer does not deviate from the list or inject any extra remarks – data from these are usually quantified. In qualitative, semi- structured or unstructured interviews, the interviewer adjusts questions to the interviewee response and may inject their own opinions and ideas – the focus is on gathering rich complex meanings.
66
Laboratory experiment
An experimental study that takes place in a controlled, laboratory environment. Although ecological validity may be low, it is easier to control for confounding variables in the laboratory than in the field.
67
Longitudinal
Involving the collection of data about an individual or group over a long period of time.
68
Lower bound
The lowest value in a confidence interval. For a 95% confidence interval, this is calculated by multiplying the standard error by 1.96 and subtracting the result from the sample mean.
69
Material data
Direct evidence from bodies and brain, for example when measuring biological reactions (e.g. hormone levels) or using brain-imaging techniques.
70
Mean
A descriptive statistic that is a measure of central tendency; an average calculated by dividing the sum of values in a data set by the number of values. The mean is usually reported if data are measured at interval or ratio level.
71
Mean deviation
A measure of dispersion used in descriptive statistics to describe
the average variation from the mean of each individual value in a data set. The mean deviation is calculated by subtracting each value in a data set from the mean
and then calculating the average. However, the standard deviation is the more generally accepted measure of dispersion reported.
72
Meaning
The content, intentions and expressions carried by the words or signs exchanged by people in communication. Also definitions, interpretations and values, a reference of equivalence or difference.
73
Measurement error
The discrepancy that may arise between scales of measurement used (e.g. for rating scales in self-reports) and the actual value of the measurements.
74
Measures of central tendency
The collective term for the descriptive statistics, mean,
| mode and median, which describe the central values, or average, of a range.
75
Measures of dispersion
The collective term for the descriptive statistics, range, mean deviation, variance and standard deviation, which describe the spread of data.
76
Median
A measure of central tendency used in descriptive statistics; an average
calculated by finding the middle value in a data set. The median is usually reported if an ordinal level scale is used, or if there are outliers in the data set (extreme scores or values).
77
Methodology
The ‘theory of methods’. It refers to the overall theoretical rationale and the principles that define how a research question, a set of methods, procedures of data collection and analysis are embedded within a theoretical framework; i.e. how they relate to the ontology and epistemology that underpins the choice of methods.
78
Methods
Techniques and tools for collecting and analysing kinds of data.
79
Mode
A measure of central tendency used in descriptive statistics; an average calculated by counting the most frequently occurring values in a data set. The mode is reported if a nominal level scale is used, giving the category with the highest frequency count.
80
Narrative
A narrative is a story which gives an account of a sequence of fictional or non- fictional events. Discourse analysts focus on the ways in which the stories are told while researchers interested in life history focus on biography.
81
Naturalistic field experiment
An experimental study that takes place outside
| the laboratory, in an environment where the variables, while still controlled, occur naturally.
82
Negative correlation
A correlation where the relationship between the two variables is positive, represented by a correlation coefficient between -1 and 0. As one variable increases in value, the other decreases.
83
Negative skew
A spread of data with more values lower than the mean than higher. The distribution is described by a non- symmetrical, bell-shaped curve which peaks to the right.
84
Nominal level scale
A number is used as an arbitrary label to distinguish between mutually exclusive and distinct groups of data, as is the case with categorical variables.
85
Normal distribution
A regular spread of data where the mean, median and mode have
similar values and which commonly occurs with natural phenomena such as height, weight or shoe size. The distribution is described by a smooth, symmetrical, bell-shaped curve with the mean at the highest point.
86
Null hypothesis
In an experiment, a formal statement that asserts that a named independent variable will not affect a named dependent variable. It states that any difference between the groups or conditions is due to sampling error. In a correlational study, it is a statement which asserts that there will be no relationship between two named variables and that any relationship observed is due to sampling error. Using statistical tests we can decide whether the data we collect allow us to reject or accept the null hypothesis.
87
Objective data
Data which are assumed to have only one interpretation and not be biased by the views of the researcher.
88
Observation
A research method where the observer notices, watches and scrutinises significant details of an event which are then recorded within a framework of previous knowledge and ideas.
89
One-tailed hypothesis
A research hypothesis which states the direction of the effect or correlation, for example whether the named independent variable will cause values measured for the dependent variable to increase or decrease. A one-tailed hypothesis should only be stated if there are extremely good grounds for making a directional prediction.
90
Ontology
Ontology is the study of, or the theory of the nature of existence (i.e. of ‘being’). In a psychological context, ontology refers to the underlying assumptions about the nature of existence as a human being. Hence, ontology can be referred to as the theory of what, in a certain perspective, is seen to constitute a human being, or, more generally, what it is to be a person.
91
Operationalise
The process of converting or defining variables of interest that allows them to be measured or manipulated.
92
Order effects
Potentially a confounding variable introduced when the same participants are allocated to more than one condition in an experiment and conditions are run, or materials are presented, in the same order for each participant. This means the effects of the independent variable cannot be isolated. Order effects may be avoided by counterbalancing, by randomising elements of the materials, or by using a between-participants design.
93
Ordinal level scale
A numerical scale where numbers are used to imply order or rank.
94
Outsider viewpoint
A viewpoint gained from using methods, such as the experiment, to collect data which can be ‘seen from the outside’ without recourse to introspections or people’s own accounts of their mental states.
95
Paired samples t-test
A parametric test of difference used in inferential statistics to analyse data produced by an experiment with a within-participants design and which has two conditions. This test may also be called the related t-test, dependent t-test, or repeated measures t-test.
96
Paradigm
A thought pattern and set of practices that define a discipline during a particular period – that is, the beliefs about what should be studied, what types of questions should be asked and how the investigations should be carried out and interpreted. In this course (DSE212) the term ‘paradigm’ is used synonymously with ‘perspective’ or ‘theoretical framework’.
97
Parametric test
Parametric tests are inferential statistical tests which make certain assumptions about the population from which the data are drawn, e.g. that the population is normally distributed. With parametric tests the data cannot have been measured using a nominal level scale. The t-tests are an example of a parametric test.
98
Participant
A person who takes part in a psychological experiment or study. Previously the term ‘subject’ was used, so you may come across this term in older publications.
99
Participant observer
A researcher who openly observes while also taking an active part in the research setting.
100
Participants’ orientations
A term associated with discourse analysis. Activities accomplished in conversation – for example, when a person says something, this utterance (orientation) sets a frame for another person’s response which shows how they have interpreted what is happening (their orientation), and which changes the frame or the context again.
101
Pearson product moment correlation coefficient
A parametric test used to analyse data from correlational studies when a linear relationship between the variables is expected. This test calculates a correlation coefficient known as ‘r’, which is considered to be the measure of effect size, or the strength of the relationship between variables.
102
Perspective
A theoretical framework that informs the types of questions asked, methods favoured and thus the types of knowledge produced.
103
Pie chart
A circular diagram used to display subsets of a whole data set, where each subset is represented as a proportion (as ‘slices of a pie’).
104
Placebo effect
The effect, particularly in drug trials, that occurs when participants in the control condition of an experiment are given a placebo (a harmless substance that looks like the drug being tested) – they still experience the apparent effects of the drug.
105
Population
The group or set of individuals or entities, defined by some characteristic(s) about which statistical inferences are to be drawn, usually based on a random sample taken from the population.
106
Positive correlation
A correlation where the relationship between the two variables is
positive, represented by a correlation coefficient between 0 and 1. As one variable increases in value, so does the other.
107
Positive skew
A spread of data with more values higher than the mean than lower. The distribution is described by a non- symmetrical, bell-shaped curve which peaks to the left.
108
Positivism/positivist model
The belief that the world as it is given to observation is the way the world actually is.
109
Practise effects
A type of order effect. A confounding variable introduced when the same participants are allocated to more than one condition in an experiment and conditions are run in the same order for each participant. Participants become more ‘practised’ as the study progresses. This means the effects of the independent variable cannot be isolated. Practise effects may be avoided by counterbalancing, or by using a between-participants design.
110
Probability
The likelihood of something happening. In research, probability is expressed numerically, so that a value of 1 equates to an event definitely occurring, and a value of 0 equates to an event never occurring.
111
Probability value
The value calculated by a statistical test as the measure of significance. It gives the level of likelihood that
differences or relationships are due to sampling error. This is often written as ‘p-value’. A probability value of .05 has been adopted as the threshold for accepting the null hypothesis (test result is greater than .05) or rejecting it (test result is equal to or less than .05) – see ‘alpha value’.
112
Qualitative analysis
Qualitative researchers critically immerse themselves in their data – they sort, organise, conceptualise, refine and interpret the data. The aim of qualitative analysis is to generate rich descriptions of common patterns and themes and theorise about how and why these relations appear as they do. Theories are contextualised in relation to the context the participants create (e.g. in what they say and do) and in relation to the research context. They are also, however, related to the context of how others have articulated the evolving knowledge in the literature.
113
Qualitative methods
Methods used to access and understand the meanings that participants give to the phenomena of interest – for example interviewing, observation, ethnography, discourse analysis, participant observation, focus groups.
114
Qualitative research
Research that describes and develops theories of how people in their natural settings understand the world and construct meanings. Explanations always refer to a specific time and context. Multiple realities and viewpoints are assumed and researchers consider themselves an explicit part of the data gathering and analysis. Reflexive analysis thus forms a crucial part of the analysis.
115
Quantitative methods
Methods that investigate psychological properties that can be counted or measured and statistically analysed. Examples include experiments, questionnaires and observational studies.
116
Quantitative research
Research that relies upon measurement and numbers, and which aims to make general statements about a population.
117
Quasi-experiment
An experimental research method where the independent variable occurs naturally (e.g. age, height, sex) and cannot be manipulated by the researcher.
118
Random allocation
The arbitrary assignment of participants to each condition of an experiment, usually introduced as a control for potential confounding variables such as individual differences.
119
Range
A measure of dispersion used in descriptive statistics, calculated by subtracting the smallest value in a data set from the largest.
120
Ratio level scale
A numerical scale where the intervals between measurements are specified and constant. Similar to an interval level scale, but where the zero point is a true zero and numbers used are not arbitrary representations.
121
Reflexivity
The process of researchers reflecting on how they may have influenced the research that was conducted, i.e. the process of identifying how a researcher’s involvement influenced, acted upon and informed the research, including the questions asked, the analysis conducted and the conclusions drawn.
122
Reliability
The extent to which a measure or test always produces the same results for an individual over time and across situations.
123
Replicability
The extent to which a study can be repeated with a different sample and still produce similar results.
124
Research hypothesis
Also known as an ‘alternative hypothesis’ or, in the case of an experiment, an ‘experimental hypothesis’. In an experiment, a formal statement that predicts that a named independent variable will have an effect on a named dependent variable. In a correlational study, a
statement which predicts there will be a relationship between two named variables. In experiments and correlational studies it is actually the null hypothesis that is tested.
125
Research question
Research questions
define and specify the focus of interest. In qualitative research, they are always embedded in a theoretical framework; i.e. they are underpinned by the epistemological and ontological assumptions of the respective research perspective and thus provide both a theoretical and practical focus to the research undertaken.
126
Sample
A subset of a population (see ‘population’).
127
Sampling distribution of the mean
The spread of a number of sample means taken
| from the same population, which will form a normal distribution if enough means are plotted.
128
Sampling error
The naturally occurring difference between the values derived from testing a sample of participants in a study and testing the entire population from which the sample was drawn. Increasing the sample size may reduce the error.
129
Scatterplot
A graph showing the relationship between two variables (represented by the axes), whereby the corresponding data values are plotted as single points. The pattern of points produced may suggest a correlation exists between the variables if they appear to fall in a straight line. The pattern of points will slope upwards from left to right for a positive correlation, and downwards from left to right for a negative correlation. Also known as scattergram or scatter graph.
130
Scientific
The rigorous, systematic and empirical study of observable behaviour using experiments, observations or correlational studies and statistical analysis.
131
Second-order coding
Following on from first-order coding, the researcher identifies codes that capture the meaning of larger segments of the data. This reduces the number of first-order codes as the data are sorted into broader and more encompassing categories.
132
Single-blind testing
In drug trials particularly, the process of not informing participants whether they are receiving the real drug or not, to avoid demand characteristics.
133
Social constructionism
A theoretical perspective that emphasises the ongoing building of self-perception and world views by individuals in dialectical interaction with society; for example, identities are not fixed and the exclusive result of biology but highly contingent on social and historical processes.
134
Standard deviation
A measure of dispersion used in descriptive statistics to describe the spread of values in a data set in relation to the mean. The standard deviation is calculated by finding the square root of the variance (see ‘variance’). It is the most accepted measure of dispersion and should always be reported if the mean is stated.
135
Standard deviation points
Related to the standard deviation of a data set, so that two standard deviation points is the value for
the standard deviation multiplied by 2, and so on. Because each standard deviation point is always at the same point on a normal distribution curve, either side of the mean, they may be used to predict the likelihood of a value. Importantly for statistical calculations, 95% of the data falls within 1.96 standard deviation points of the mean.
136
Standard error
The standard deviation associated with the sampling distribution of the mean. The standard error is calculated by dividing the standard deviation of a sample by the square root of the sample size. Calculating the standard error allows the population mean to be calculated within a 95% confidence interval.
137
Standardised measure
A measure that is the same no matter how large or small the numbers involved. For instance, the statistic d is a standardised measure of effect size, calculated by subtracting one sample mean from the other and dividing the result by the standard deviation. This gives a more meaningful result that can be compared across studies.
138
Statistical significance
In inferential statistics, if it is found that the results are not due to sampling error, then the null hypothesis can be rejected and the results (either the difference between conditions or the relationship between the variables) can be said to be statistically significant. A probability value of .05 has been adopted as the threshold for accepting the null hypothesis (test result is greater than .05) or rejecting it (test result is equal to or less than .05) – see ‘alpha value’.
139
Structured observation
The noting and recording of the frequency of behaviours, usually subject to statistical analysis.
140
Subjective data
Data based on or influenced by personal experiences, thoughts, feelings, opinions.
141
Subject positions
A term associated with discourse analysis, but used more widely. The identity, opinion or stance that is constructed when a person draws on a particular interpretative repertoire from the stock available in their culture.
142
Superordinate constructs
Superordinate (or super-ordinate) constructs subsume lower-order codes or constructs. In order to do this, they tend to be more abstract or conceptual than lower-order constructs and to bring together two or more lower-order constructs. For example, rain, sleet and biting wind are simple descriptions that
can be subsumed into ‘bad weather’ or ‘dramatic weather’ or ‘brooding day’. All of these are superordinate constructs of the description.
143
Symbolic data
Symbolic creations of minds, such as texts or art.
144
Thematic analysis
The process where qualitative data, for example interview transcripts, are reorganised into categories and themes.
145
Themes
Recurrent topics, ideas and statements that form patterns which can be brought together into a category.
146
Theory
A logical model or framework for describing or conceptualising the behaviour of a related set of natural or social phenomena.
147
Third-order coding
The identification of superordinate constructs – the themes – which capture the overarching meaning of second-order codes and so are necessarily broader. The process is also known as pattern coding since the overarching themes represent more general concepts than second-order codes and often incorporate several lower-level codes. This drawing out of overarching themes and the identification of quotations that exemplify them is the final part of the process of thematic analysis.
148
t-tests
Parametric tests of difference, used in inferential statistics to determine whether the difference between two conditions in an experiment is statistically significant. t-tests produce a statistic referred to as ‘t’ which compares the variation between conditions (i.e. the difference between the means) and the variation within the conditions (i.e. the standard deviation for each). The particular t-test used depends on the design of the experiment.
149
Two-tailed hypothesis
A research hypothesis which states a difference or a relationship but does not state a particular direction, for example that the named independent variable will affect the dependent variable to either increase or decrease the values.
150
Type 1 error
Where the null hypothesis is rejected incorrectly when it should have been accepted.
151
Type 2 error
Where the null hypothesis is accepted incorrectly when it should have been rejected.
152
Unstructured naturalistic observation
A qualitative approach committed to avoiding disturbance of the activities that are being observed.
153
Upper bound
The highest value in a confidence interval. For a 95% confidence interval, this is calculated by multiplying the
standard error by 1.96 and adding the result to the sample mean.
154
Validity
The extent to which a measure or test measures what it claims to.
155
Variables
Things or properties that we can measure and which have values that can vary; for example, attributes (e.g. attractiveness) and characteristics (e.g. height).
156
Variance
A measure of dispersion used in descriptive statistics to describe the mean of squared deviation values (calculated by subtracting the mean from each value and squaring the result – squaring avoids any negative values). It is usually used as a step on the way towards calculating the standard deviation. Variance is calculated by dividing the sum of squared deviation values in a data set by the number of values.
157
Voice
One focus of discourse analysts is the different voices that people bring into their narratives. According to the linguist Bakhtin, when people speak their narratives are populated with the voices of others – typically those who are significant to the person; for example, parents, teachers, friends, partners, people who are
influential.
158
Within-participants design
An experimental design which uses the same participants in each condition, so fewer participants are needed than for a between-participants design. Also known as paired samples
design or repeated measures design. Can reduce the effects of individual differences, but those of demand characteristics and/or order effects have to be taken into consideration.
