Metod och Analys II Flashcards
1
Q
Descriptive statistics – three important parts
A
- Frequency distribution
- Central tendency
- Variability
2
Q
Descriptive statistics – three purposes
A
- Determining how many people got each score
- Providing information on the standing of a score relative to all other scores
- Graphically summarizing the set of scores
3
Q
What could be 4 purposes of Frequency distribution?
A
- It is a record of the number of people with each score (or in each category) off the variable
- It allows examination of the full distribution at a “glance”
- Ideally, this will allow the reader to get a basic understanding of the data without being overwhelmed by all the raw scores
- It provides a visual assessment of central tendency and variability
4
Q
What is two important factors when it comes to frequency distribution?
A
- There should be a listing of each possible score and the frequency occurrence
- As a check, the sum of the frequencies should be equal to n (the sample size)
5
Q
There is the characteristics of frequency distribution shapes, describe them.
A
- Modality: the number of humps in a distribution
- Skewness: is a measure of whether the distribution is symmetrical or not
- Kurtosis: characterizes the relative peaked-ness/flatness of a distribution compared to the normal distribution
6
Q
What is Normal Distribution?
A
- Can be described as the bell-shaped curve
- The majority of scores lie around the center of the distribution
- Symmetrical curve
7
Q
What is left (negative) skewed distribution?
A
- Frequent scores are clustered at the higher end & tail points towards the lower negative scores
- Not symmetrical curve
8
Q
What is right (positive) skewed distribution?
A
- Frequent scores are clustered at the lower end & tail points towards the higher or more positive scores
- Not symmetrical curve
9
Q
What is Leptokurtic distribution?
A
- The curve is symmetrical, similar to a normal distribution
- But the center peak is much higher
- That is, frequent scores are near the mean
10
Q
What is Platykurtic distribution?
A
- The curve is symmetrical, similar to a normal distribution
- But the frequency of most of the values are the same
- As a result, the curve is very flat, or plateau-lake
11
Q
What happens when data are not normally distributed?
A
- Data that are positively skewed (many scores are low) may cause the mean score to be artificially inflated
Resulting in the mean pushed to higher a higher score - Data that are negatively skewed (many scores are high) might lead to an artificially deflated mean
Resulting in the mean pushed to a lower score - Leptokurtic distributions (high peak) may offer little variation in the data
Resulting in risk to not detect result - Platykurtic distributions (low peak) may offer too much variation in the data
Resulting in risk to have too high results - IF normal distribution has been compromised, we may have less confidence in the outcome of parametric tests
12
Q
How do you measure normal distribution in SPSS?
A
- You have to check for histograms
- If skewness and kurtosis values are -/+ 1 range, we can assume that distribution is normal – strict criteria
- If skewness and kurtosis values are -/+ 2 range, we can assume distribution is normal – reasonable criteria
- When assessing statistical normal distribution – we use Kolmogorov-Smirnov test if the N is larger than 50
we use the Shapiro-Wilk test if the N is smaller than 50
13
Q
What can we do if the distribution happens to not be normal?
A
- Check for outliers
- Transform data
- See textbook 57-61
14
Q
What is Central Tendency?
A
- The goal of central tendency is to describe the average score on a variable for a distribution (eg sample or population)
- Ideally, this will be a single value, this will be an estimate of the middle or typical score in the distribution
15
Q
Which three common measures is there to measure central tendency?
A
- Mean – problable the measure most frequently thought od as the average
- Median – the middle score merely as a function of the total number of scores in the distribution
- Mode – the most frequently occurring score in distribution
16
Q
In which 3 ways are distribution shapes and central tendency measures correlated?
A
- In a perfectly normal distribution, the mean, median and mode are the same value
Mean = median = mode - In a positively (right) skewed distribution the mean is bigger than the median and the meadian is bigger than the mode
Mean > median > mode - In a negatively (left) skewed distribution the mean is smaller than the median and the median is smaller than the mode
Mean < median < mode
17
Q
What is variability?
A
- Variability refers to how spread out the scores in the distribution are
- The mean is good for representing the typical score of a distribution, but the mean alone does not completely describe the distribution
- For example – two different distribution both has the sample size n = 1000, and each has the mean M = 100
But we still know nothing how the scores are spread out
18
Q
What is variability?
Which tree ways of measuring variability are there?
A
- Range
- Interquartile range
- Standard deviation (most frequently used)
19
Q
What is Standard Deviation?
A
- A deviation score is merely the difference between individual score (Xi) and the mean of the distribution (e.g. M)
- Deviation score = (Xi-M)
- We can think of the standard deviation as an average deviation score
- For example – we would expect smaller deviation scores, on average, in a distribution that has less variability (spread in the scores)
20
Q
What is Standard Deviation?
What is the Standard deviation in a normal distributed sample?
A
- 68% falls within 1 standard deviation from the mean
- 95% falls within 2 standard deviations from the mean
- 99,7% falls within 3 standard deviations from the mean
21
Q
How can inferential statistics be described?
A
In opposite to descriptive statistics we no longer try to describe our sample, we now try to imply/inference the statistics on the population
22
Q
What is a direct vs an indirect approach when it comes to hypothesis testing?
A
Direct approach
* Conduct the study in the entire population
* Determine if the hypothesis is supported
* Is typically not feasible or even possible
Indirect approach
* Obtain a sample from the population
* Compute statistics in the sample (e.g. mean)
* Infer relations in population from the sample
23
Q
There are 2 different types of hypotheses, these are?
A
- Scientific hypothesis
- Statistical hypothesis
- Null hypothesis
- Alternative hypothesis
24
Q
What is the Scientific hypothesis?
A
This is what the researcher expects to find
Eg
* A new type of therapy will be more effective at reducing depressive symptoms that the old type
* Depression is related to low life satisfaction
25
Describe the two different types of Statistical hypotheses
Null hypothesis
* Symolized: H0
* The hypothesis of “no” effect
Ex. If testing for mean differences between two groups, H0 would specify no difference is present (no effect)
Alternative hypothesis
* Symolized: H1
* Hypothesis of “effect”
Ex. Testing for mean differences between two groups, H1 must specify all other outcomes other than 0
26
What is the p value?
* We assume that the null hypothesis is true (e.g. no effect)
* We fit a statistical model to our data that represents the alternative hypothesis
* We calculate probability of getting that model if the null hypothesis were true
* If the probability is very small (p < 0.05) we conclude that the model fits the data well – we reject the null hypothesis and gain confidence in the alternative hypothesis
* In other words, we conclude that the likelihood of getting our findings by chance is less than 5%
* As it is the null that is tested, process often referred to as “null hypothesis testing”
27
In reality H0 is either true or not true. Thus, there are four possible outcomes of statistical inference – which are these?
1. The null hypothesis is correctly retained
The null hypothesis is true and not rejected
2. Type I error (α) – false alarm
The null hypothesis is true but rejected
3. Type II error (β) – missing the effect
The null hypothesis is false and not rejected
4. Correct rejection of the null hypothesis (1 – β e.g. power)
The null hypothesis is false and rejected
28
What is one-tailed vs. two-tailed tests?
* One-tailed hypothesis has a specific directional prediction
(e.g. patients depression scores will decrease after undergoing therapy)
* Two-tailed hypothesis are non-directional predictions
(e.g. there will be a difference among males and females in their life satisfaction level)
* In this course – we will only be using the two-tailed hypothesis and two-tailed tests
29
What is effect size – and what are the 3 types called?
Effect size is the actual magnitude of the difference between groups or the magnitude of the association between variables
1. Cohen´s d (can exceed 1)
Used when the mean difference is tested
*Small cohen´s d is <0.25
Medium cohen´s d is 0.25 - 0.4
Large cohen´s d is 0.4 - ∞ *
2. Pearson´s r (ranges from 0-1)
Used when correlation is tested
3. Eta-square
Used when variance is tested
30
In inferential statistics we have two different types, what are they called?
* Parametric tests
Is used when the outcome variable is continuous
(and will be the focus in this course)
* Non-parametric tests
Is used when the outcome variable not is continuous
31
Which are the 4 basic assumptions that may be applied to most of the parametric tests?
1. Dependent variable is normally distributed
2. Homogeneity of variance
3. Outcome variable is continuous (interval or ratio)
4. Independence of observations
32
How can we know if the dependent variable is normally distributed?
* Check skewness and kurtosis values
* Check histogram
* Check normality by conducting some tests (Kolmogorov-smirnov or the Shapiro-Wilk test)
If the resulting p-value is under 0.05, then we have significant evidence that the sample is not normal, so we´re “hoping for a p-value of 0.05 or above.
33
What is homogeneity of variance?
* The variance of a variable should be stable throughout different levels of another variable
* Assume that you developed a delinquency prevention program. You randomly assigned a group of youth to prevention program, you compared the two groups on their engagement in delinquent behaviors (conducted independent sample t-test)
* The variance of delinquent behaviors in the prevention and control group should be roughly the same (Levene´s test)
34
What does independence of observations mean?
Data that is obtained from different Xi (participants) are independent
35
What does it mean that outcome variable is continuous?
* The outcome variable in parametric tests should be continuous
* In other words, the outcome variable should be measured on interval or ratio scale
36
What is t-tests?
T-tests are used to compare the means of two groups on a given variable
Two different types of t-test:
* Related t-test
* Independent t-test
37
What is related t-test?
* Examines difference in mean dependent variable scores across two within-group conditions (independent variable), measured across a single group
* Each (and every) participant in group 1 can be paired with a participant in group 2
* Participants can be matched/paired with themselves
* Matching is on a consistent basis (same rule for matching is applied to each pair)
* The purpose of using paired samples design is to reduce data variability (to reduce error in the outcome variable)
38
What is the t-value?
* Given that the obtained t-value is smaller than the critical t-value, retain the null hypothesis
* Simply, the results tell us that there is no apparent reason to believe that (for example husbands and wives) differ in their ratings
* Given that the obtained t-value is larger than the critical t-value, reject the null hypothesis
* Simply, the results tell us that the two groups differed significantly.
39
What is the inter item reliability measure called and how do we know if it´s acceptable?
Cronbach´s Alfa
The scale goes from 0 to 1.
In most cases above 0.7 to 0.8 is considered acceptable.
40
What is independent samples t-test?
* If one wants to compare two independent groups on a given variable, then independent samples t-test should be used
* The groups are assumed to be independent when there is no consistent basis on which the groups are matched
* Simply, participants in one group are not “related” in any way to participants of the other group
41
When is it more optimal to use analysis of variance (ANOVA) instead of t-test?
* When we want to compare more than two groups
42
Why is it risky to analyze the data as a series of t-test when having more than two groups?
The approach will result in an inflated type 1 error rate
(e.g. probability of rejecting the null hypothesis when the null hypothesis is true - to assume that there is a significant difference where it´s not)
43
If we would like to do multiple t-tests to compare the means of multiple groups, what would we have to do and why?
* If we run many comparisons back to each other, we would increase the probability of finding a significant difference by chance
* Therefor – we have to divide 0.05 with the number of group comparisons
* If we have 10 comparisons, for each comparison, our new p-value must be p < 0.005 so that the study p value would remain p < 0.05 in the study
* To adjust the p value for multiple group comparisons is not a practical approach
44
What is tested in one way ANOVA?
To test the mean differences across multiple groups on a given variable
45
What is the logic behind ANOVA?
* When more than two means are involved, it is not possible to calculate a single value that would represent the differences in means between the groups
* Instead, an estimate of variance among the group means is used
If the group means are fairly similar, this should result in small variance
Large differences among the means should result in larger variance
* We try to understand whether there is a significant difference across different levels of an independent variable on the level of an outcome variable
46
Which 3 reasons exist to explain between-group variance (e.g. the means of different groups in same sample differ)
1. Individual differences
2. Experimental error
3. Systematic reason (e.g. treatment effect in this sample)
47
What is individual differences (in between-group variance)?
For example, participants may come into the experiment with different histories
(e.g., some may be better able to cope with stress, some may be better at the performance task)
48
What is experimental error (in between-group variance)?
Whenever a measurement is made, the instrument is subject to small amounts of error
(e.g., misunderstanding the participant´s verbal response)
49
What is systematic reason (in between-group variance)?
The different conditions of treatment causes the groups to be different
50
Which 2 reasons exist to explain within-group variance (e.g. mean difference in a given group)?
1. Individual differences
2. Experimental error
51
When do we reject the null hypothesis considered the f-value (when calculated by hand)?
* Go and find the critical F-value from a F-distribution table
* If the f-statistics that you find is smaller than the critical F-value, than you retain the null hypothesis (i.e., there is no difference across groups)
* If the f-statistics that you found is larger that the critical F-value, then you reject the null hypothesis (i.e., there is a difference across groups)
52
What do a significant f-value tell us?
Whether there is a statistically significant difference across groups (But not which group is different than the other one)
53
Which are the 2 options I can choose between if I want to know which group that is significantly different from the other one? (When it comes to one way ANOVA)
1. Planned contrasts
Is used when you have a specific hypothesis
For example: Group A shows a bigger difference than group B
2. Post hoc tests
Is used when you have no specific hypothesis
For example: I don’t know where the difference is, but I want to know
54
When do we use factorial, and when do we use one-way ANOVA?
Factorial: when we have two or more independent variables
One way: when there is one independent variable, which is measured on a nominal scale, and has **two or more** levels
55
What are we interested in when we use factorial ANOVA?
In the main effects of independent variables as well as their interaction effect on the dependent variable
56
What does interaction effect mean?
Interaction effect occurs when the effect of one independent variable on the outcome variable depends on the level of another independent variable
57
Can you describe what 2x3 factorial design means?
For example if a study design is:
* IV 1 = interviewers’ gender
(two levels: female and male)
* IV 2 = interviewers’ eye contact
(tree levels: low, moderate, high)
* DV = liking
The 2 means 2 levels in the first IV, and 3 means three levels in the second IV
58
What is important considering IV in factorial ANVOA?
That IV is categorical
For example gender
59
There is three questions that’s possible to answer with help from factorial ANOVA, which?
1. Is there a significant main effect of IV1 on DV?
For example: do interviewees perception about interviewers differ based on interviewers gender?
2. IS there a significant main effect on IV2 on DV?
For example: do interviewees perception about interviewers differ based on the level of interviewers eye contact?
3. Is there a significant interaction effect of IV1 and IV 2 on DV?
For example: does the effect of interviewers gender on interviewees perception vary depending on the level of interviewers eye contact?
60
Why do we use Repeated measures ANOVA?
* When the same participants are followed over time
OR
* They attend all the conditions in an experiment
* In these cases, we want to examine whether people change over time, in mean level of a given variable, we use repeated measures ANOVA
61
Can you give an example of a research design where the participants attend all the conditions in an experiment
* a researcher is interested in understanding whether teaching methods of a research design course influence students ‘course satisfaction
* She samples a group of 50 C-level Psychology students
* First, she puts these students into a condition where the professor solely lectures for 2 hours. After the lecture, she administers a 10-item survey assessing students’ course satisfaction
* A week later, she puts the same students in another condition where the professor integrates lecture and interactive student activities together during the course. Then, she administers the same course satisfaction survey to the students
* Finally, she compares the students’ course satisfaction across the two conditions
IV = teaching methods
IV has two levels, which are “solely lecture based ”and “integration of lecture and interactive student activities”
DV = course satisfaction
The participants attended all the conditions in the experiment
62
Can you give an example of a research design where the same participants are followed over time
* A researcher wanted to know whether A, B, or Students spend more time for studying
* She contacted asked all the psychology A students to report how many hours they spend for studying per week in 2013
* Assume that all the A students passed their courses, and enrolled in B courses a year later
* In 2014, Sofia asked B students to report how many hours they spend for studying per week
* In 2015, she asked C students to report how many hours they spend for studying per week
* After the data collection was completed, she examined whether the average amount of time that students spent for studying changeover time
IV = students’ class standing IV has three levels, which are “A students,” “B students,” and “C students”
DV = hours spent for studying per week
Participants are followed over time!
63
When do we use correlation?
* When we want to measure the relation between two variables
* The form of the relation is linear (“straight line”)
* How related are two variables, as one variable changes (varies), does the other variable change in a systematic fashion?
* I.e. do they covary?
64
What is Pearson Product Moment Correlation (PPMC) coefficient?
* The most general, and typical type of correlation
* Is used when both variables are continuous
* Sign for this is a small “r”
65
What does the two directions positive and negative mean?
* A positive direction (reported just as they are) means that higher numbers on X is associated with higher numbers on Y
* A negative direction (reported with a “-“ in front of the numbers) means that higher numbers on X is associated with lower number on Y
66
What does magnitude refer to, and how is it indicated and what does the correlation of 1.00 and 0.00 mean?
* How much (or how well) the two variables are related
* Indicated by a numeric value of correlation (the sign is not relevant)
.45 and -.45 has the same magnitude
* 1.00 = means perfect correlation
* 0.00 = no correlation
67
What are the guidelines for magnitude?
* .10 = small
* .30 = moderate
* .50 = large
68
What are 3 other types of correlation, and when you use them?
* Spearman rank order correlation
When one, or both of the variables are measured on the ordinal (rank) scale
* Point biserial correlation
When one variable is continuous and one variable is dischotomous (two levels)
* Phi coefficient
(When both variables are dischotomous (two levels)
69
When is partial correlation used?
* When association between two variables while controlling some other factors (eg confounding variables)
* Is interested in understanding the association between two variables after controlling for another variable
* Ex: interested in association between shyness and loneliness among children, after controlling for child’s gender
Variables of interest: shyness and loneliness
Variables controlled for: child gender
70
When do we use regression analysis?
* When we have a prediction
* Ex: does taking notes during the lecture (predictor variable) predict success in the exam (outcome variable)
* But still linear association
71
How do we know if a variable is a predictor variable or an outcome variable?
* This is a theoretical question, and we need theoretical justification
* It needs to rely on a theory and conceptual arguments, when determining which variable will be a predictor and which one will be an outcome variable
72
What does regression analysis provide us?
* Unique effect of predictors on the outcome variable
* Again, correlation only provides us the bivariate association between only two variables
* One can examine the unique effects of multiple predictors variables on an outcome variable in a regression analysis
* Ex: Do parents’ overprotective behavior or adolescents’ socially anxious behaviors predict feelings of loneliness among youth?
73
When is simple/univariate regression model used?
* When one examines the effect of one predictor on an outcome variable
* Outcome variable should be a continuous variable (but no restriction on the predictor variable)
* Ex: dies experiencing stress at work predict individuals work satisfaction
74
When is multiple regression model used?
* When one examines the effect of more than one predictor variable on an outcome variable
* Can estimate the effect of each predictor variable on the outcome variable, above and beyond the effect of other predictor variable in the equation
* Outcome variable should be a continuous variable (predictor variable doesn’t need to be continuous)
75
What is multicollinearity and how is it interpreted?
* Overly high correlations between the predictor variables
* Check multicollinearity before running a regression model (or before interpreting it at least)
* Ideally the predictor variables should not correlate to each other, but some association could be tolerated
* When multicollinearity arise, the regression estimates would be biased
* Tolerance – should be above .10 (or .20)
Values less than .10 reflect severe multicollinearity – reflecting that 90% of the predictor in question is redundant with the remaining predictors (how much overlap we have) – the closer to one the better
* VIF (variance inflation factor) – 5 or lower
Tells how much variance of an estimated regression coefficient is increased due to collinearity. VIF is merely the reciprocal of tolerance and does not provide any unique information with respect to multicollinearity in the predictors
76
What can we do if we have highly correlated predictors? (i.e. multicollinearity problem)
* Remove one predictor
* Combine the correlated predictors (if it’s theoretically applicable)
* We can use latent variable modelling (which is not covered by the course)
77
What is factor analysis?
* Used to estimate whether all the items in a scale belong to the same construct
(are the indicators driven by the same underlying factor)
* Can also be used to see how a scale works in another context
(for example if we want to bring a measure to a Swedish context from a German context)
78
What are three types of factor analyses?
* Principal component analysis (PCA)
* Exploratory factor analysis
* Confirmatory factor analysis
79
What is important to conduct before running a factor analysis?
A correlation between the scale items, we need at least reasonable correlation between the items
80
What do we do if an item is high in more than factor?
* Must ask myself – is this item important
* If yes - See where it conceptually belong
* If no – take it out
81
Name some (5) differences in quantitative and qualitative research
1. Scientific emphasis
Quantitative: confirmation and falsification, focuses on testing hypotheses and theories
Vs
Qualitative: exploratory, focuses on generating hypotheses and theories
2. Research objectives
Quantitative: explain (cause and effect), control, predict, description of characteristics of population
Vs
Qualitative: explore, particular description, depth of understanding, social construction of reality
3. Data
Quantitative: quantitatively measured variables (numbers)
Vs
Qualitative: words, text, images, documents
4. Results
Quantitative: generalizable findings
Vs
Qualitative: particularistic findings and claims
5. Final report
Quantitative: statistical results (with significance testing of correlations, differences between means) with discussion of results
Vs
Qualitative: narrative with rich contextual description and many direct quotations
82
Name some (5) differences in quantitative and qualitative research
Name some (3) different qualitative methods
* Interview methods
Data: responses to interview questions
To make sure it’s reliable: use an interview protocol
* Qualitative surveys
Data: responses to open-ended question, written reflections on a vignette or story completion
* Observational methods
Data: Video recording od a setting or notes of a observer about behaviors of people in a setting
83
What is important to think of as interviewing participants in a qualitative research study?
* Ask open questions
* Ask non-leading questions
* Ask singular questions
(don’t ask about multiple things)
* Ask short, clear and precise questions
* Ask Linguistically appropriate questions
* Ask non-assumptive questions
84
What’s (9) things to thing of when interviewing?
* The physical space
* The distance
* Non-judgmental attitude
* Recording
* Use of probing questions
* Keeping focus
* Developing trust
* Participants distress and reluctance to answer questions
* Not letting the interview to turs into a conversation
85
What´s important to think of before an interview?
* Use an interview protocol
- Clear structure
- Ordering the questions
- Language check for accuracy and level
- Testing the protocol
* Training the interviewers
- Standardization of the interview process
- Ensuring quality across interviewers
86
Name some (6) different ways to analyze qualitative data
1. Thematic analysis
(the most common)
2. Grounded theory
3. Discourse analysis
4. Interpretative phenomenological analysis
5. Conversation analysis
6. Narrative analysis
87
Describe thematic analysis, which types exist and what is the specific description of each type?
Description: a method for identifying themes and patterns of meaning across a dataset in relation to a research question. Most widely used qualitative method of data analyses.
Inductive TA:
* Aims to generate an analysis from the bottom up
* Analysis is not shaped by existing theory (but is shaped by researchers’ standpoint, disciplinary knowledge)
Theoretical TA:
* Analysis is guided by an existing theory and theoretical concepts (as well as is shaped by researchers’ standpoint, disciplinary knowledge)
88
Name some advantages to thematic analysis
* Flexibility in terms of theoretical framework, research questions, methods of data collection and sample size
+ It works with a wide range of research questions, from those about people’s experiences or understandings to those about the representation and construction of particular phenomena in particular contexts
+ It can be used to analyze different types of data, from secondary sources such as media to transcripts of focus groups or interviews
+ It works with large or small data-sets
+ It can be applied to produce data-driven or theory-driven analyses
* A great “starter” qualitative method
+ Accessible for researchers with little qualitative research experience
89
There is six phases of thematic analysis, what is the first?
1. Familiarize yourself with your data“...the researcher must immerse themselves in, and become intimately familiar with, their data; reading and re-reading the data (and listening to audio-recorded data at least once, if relevant) and noting any initial analytic observations.” (Clarke & Braun, 2013)
90
There is six phases of thematic analysis, what is the second?
2. Generate initial codes
* “...involves generating pithy labels for important features of the data of relevance to the (broad)research question guiding the analysis.” (Clarke &Braun, 2013)
* “Coding is not simply a method of data reduction, it is also an analytic process, so codes capture both asemantic and conceptual reading of the data.”(Clarke & Braun, 2013)
91
There is six phases of thematic analysis, what is the third?
1. Search for themes
* Review the codes and collapse them into coherent and meaningful groups (themes)
* “Searching for themes is a bit like coding your codes to identify similarity in the data. This ‘searching’ is an active process; themes are not hidden in the data waiting to be discovered by researcher, rather the researcher constructs themes” (Clarke & Braun,2013)
92
There is six phases of thematic analysis, what is the fourth?
4. Review the themes
* “...reflect on whether the themes tell a convincing and compelling story about the data” (Clarke & Braun, 2013)
* Review the consistency in each theme
A theme includes multiple distinct phenomena – think of splitting it into multiple themes
* Review the relations across themes
Too much overlap across themes (?) – think of collapsing them
* Reflect on whether the constructed themes are providing an answer to the research question – if not, disregard the theme
93
There is six phases of thematic analysis, what is the fifth?
1. Define and name the themes
* Generate clear definitions and names for each theme
* HOW?
What story does this theme tell?
How does this theme fit into the overall story about the data?
* “..identify the ‘essence’ of each theme and construct a concise, punchy and informative name for each theme”(Clarke & Braun, 2013)
v
94
There is six phases of thematic analysis, what is the sixth?
6. Write the report
* Tell a coherent study about the data to your audience
* HOW?
+ Clearly define codes and themes
+ Present quotes that represent the codes and themes
+ Contextualize the results in relation to the existing literature
95
Name three important things about who is conducting the thematic analysis
* At least two persons collaborate in conducting the analyses
+ All analysis could be done by two independent persons
+ A subset could be done together
+ A subset could be done by both and consistency could be checked
* Agreement between the raters – inter-rater agreement, kappa coefficient
* A clear plan on how to handle disagreements
96
What are some (4) pitfalls to avoid when doing thematic analysis?
* Defining the interview questions or quotations from data as the ”themes”
* Creating themes that overlap to a greater extent ̈
* A mismatch between the data and the analytical claims ̈
* A mismatch between research questions and the form of thematic analysis
