Associate Professor Jeffrey Fuller

1
Data Management Analysis

• Associate Professor Jeffrey Fuller
• (slides adapted from Francis Boreland Gaynor
  Heading)
• University Department of Rural Health (Northern
  Rivers)

2
Learning objectives

• Data entry cleaning/ verification
• How to store, manipulate retrieve data to see
  findings
• Drawing conclusions

3
Data management

• Keep files in order
• Data collection manual (team)
• Recording procedures
• Electronic storage

4
Data cleaning or verification

• Cleaning Verification
• Quantitative Qualitative
• Frequency check Member checks
• Missing values Return transcript
• Extreme values Feedback summary
• Consistency check
• Logical response flow

5
Quantitative Analysis

• Key Questions
• How frequent is the event?
• How strongly associated with suspected cause?
• Likely to be real or due to chance?

6
Data entry case by variable format
7
Data analysis one way table of frequencies
8
Data analysis cumulative frequency distribution
9
Data analysis bar chart
10
Data analysis pie chart
11
Data analysis two way table of frequencies
12
Did the treatment work?
INTERVENTION HEART DISEASE TOTAL Present
Absent Yes 20 1980 2,000 No 32 1968
2,000 TOTAL 52 3948 4,000 Incidence yes
20 / 2000 10/1000 Incidence no 32 / 2000
16/1000
13
How strong was the effect?
Relative risk Derived from estimates of
incidence so can only be directly estimated from
RCT cohort studies. RR Incidence in exposed
group/ Incidence in unexposed group Incidence
yes 10/1000 RR 10/1000 0.625 Incidence
no 16/1000 16/1000
14
Interpreting the strength?
15
Real or chance difference?
How do we decide whether the difference reflects
a real difference between the treatments, or just
the usual variation we would expect between
samples drawn from the same population?
16
Can never know for certain
But we can infer, based on three pieces of
information 1. The likelihood the two groups are
really different. 2. The size of the
difference. 3. The precision with which the size
of the difference is known.
Need a statistician
17
Probability p value
P 0.05 95 times in 100 we will correctly
conclude there is a difference between the
groups BUT 5 time in 100 we will think there is a
difference when there really isnt We can never
know which are the 5 times!
18
Probability p value
19
Probability confidence interval
95 confidence interval range of values in
sample which probably contain the real value in
the population. Difference in RR if CI ? 1 RR
heart disease in treatment vs control 0.625
(95 CI 0.375 0.875)
20
Qualitative Analysis

• Approaches to analysis
• What is coding? how it relates to thematic
  analysis
• How to work with transcripts
• The researchers role
• Code books

21
Approaches to analysis 1 - sampling

• Inductive approach
• Involves theoretical sampling
• Breadth or depth
• Negative deviant cases
• Constant comparison
• comparing one piece of data to another
• comparing data relevant to each category (theme)

22
Approaches to data analysis 2

• Descriptive account (thin)
• Analytic account description (thick)
• Context
• Intentions meanings
• Evolution or consequence
• Classification
• Conceptual framework
• Connections
• Regularities relationships

23
Approaches to data analysis 3

• Coding, sorting and organising data
• Thematic analysis
• Often code from data inductive
• Name incident, idea, event
• Content analysis
• Develop prior categories display number of
  occurrences with who

24
What is coding

• Classifying data
• Categories
• Code label
• Related to themes
• Code to build understanding what is the data
  saying?
• Data Reduction
• Text is fractured
• Collate all chunks coded in the same way
• Pile sort

25
Types of codes

• base codes
• Gender 1
• Male 1/1
• Female 1/2
• Treatment 2
• HRT/no 2/1
• HRT/yes 2/2

• concept codes
• In vivo/ evaluator/ or predetermined codes
• Code label reflects theme
• Short is best (Nickname)
• Theme Financial barriers
• Code FinBarr

26
I was so terrified. I didn't know what to expect.
I was so ashamed that I was going to a loony
bin. I thought everybody would be mad. But
then I saw my friend Ann. She smiled and said
hello and she started asking me about the kids.
It was good to see someone I knew.
 
27
Reading interpretively Reading through or beyond
the data

• Focus on
• Respondents interpretations and understandings
• Own interpretation
• Or both respondents / own interpretation
• Involves constructing/documenting a version of
  what you think the data mean or represent
• What you can infer from the data?

28
When to commence analysis?

• At the beginning
• In the field
• Literature
• Develop a theoretical orientation

29
Coding tool - codebook
30
Analysis - considerations

• Theory / literature links
• Member checking (validity)
• Resources (depth of analysis)
• Premature closure

31
Resources

• Gifford, S. (1998) Analysis of non-numerical
  research in Handbook of Public Health Methods,
  Kerr, Taylor Heard (eds)., McGraw-Hill, Sydney,
  pp 543-554.

32
transcript