Analyzing Data

Data Analysis
 The process through which data are organized so that
comparisons can be made and conclusions drawn
Presenting Your Data
 Describe your methods – tell your audience exactly
what it is that you did to come up with your results.
 Be descriptive & analytical (so conclusions are
accepted)
 Use lots of quotes (as evidence, let people see them)
 Consider tables, charts, figures, models & diagrams
 Contextualize – where does this data come from, who
does it apply to (particularity, applicability)...
 Speak with confidence about your data
Data Analytic Strategies
 Recursive analytic strategies:
analyze cases generate findings
draw conclusion from grounded theory write
report
Analyzing Survey Data
Do you want to report…
 how many people answered a, b, c, d?
 the average number or score?
 a change in score between two points in time?
 how people compared?
 how many people reached a certain level?
Common descriptive statistics
 Count (frequencies)
 Percentage
 Mean
 Mode
 Median
 Range
 Standard deviation
 Variance
 Ranking
Getting your data ready
 Assign a unique identifier
 Organize and keep all forms (questionnaires,
interviews, testimonials)
 Check for completeness and accuracy
 Remove those that are incomplete or do not make
sense
Data entry computer screen
Smoking: 1 (YES) 2 (NO)

Survey Q1 Do you Q2 Age Q3 Support

ID smoke? ordinance?
001 1 24 2
002 1 18 2
003 2 36 1
004 2 48 1
005 1 26 1
 Supports Opposes Undecided/
restaurant restaurant declined to
ordinance ordinance comment

Current 8 33 14
smokers (15% of (60% of (25% of
(n=55) smokers) smokers) smokers)

Non-smokers 170 16 12
(n=200) (86% of non- (8% of non- (6% of non-
smokers) smokers) smokers)

Total 178 49 26
(N=255) (70% of all (19% of all (11% of all
respondents) respondents) respondents)
Summarizing data
 Tables
 Simplest way to summarize data
 Data are presented as absolute numbers or percentages
 Charts and graphs
 Visual representation of data
 Data are presented as absolute numbers or percentages
Basic guidance when summarizing data
 Ensure graphic has a title
 Label the components of your graphic
 Indicate source of data with date
 Provide number of observations (n=xx) as a reference
point
 Add footnote if more information is needed
Tables: Frequency distribution

Set of categories with numerical counts

Year Number of births

1900 61
1901 58
1902 75
Tables: Relative frequency

number of values within an interval

total number of values in the table x 100

Year # births (n) Relative frequency (%)

1900–1909 35 27
1910–1919 46 34
1920–1929 51 39
Total 132 100.0
Tables

Percentage of births by decade between 1900 and 1929

Year Number of births (n) Relative frequency (%)

1900–1909 35 27
1910–1919 46 34
1920–1929 51 39
Total 132 100.0

Source: U.S. Census data, 1900–1929.

Charts and graphs
 Charts and graphs are used to portray:
 Trends, relationships, and comparisons
 The most informative are simple and self-explanatory
Use the right type of graphic
 Charts and graphs
 Bar chart: comparisons, categories of data
 Line graph: display trends over time
 Pie chart: show percentages or proportional share
Bar chart
Comparing categories

4
Site 1
3
Site 2
2 Site 3

0
Quarter 1 Quarter 2 Quarter 3 Quarter 4
 Percentage of new enrollees tested for HIV at each site,
by quarter

6
% o f new enrollees tested for

5
4
3
HIV

2
Site 1
1 Site 2
0 Site 3
Quarter 1 Quarter 2 Quarter 3 Quarter 4
Months

Data Source: Program records, AIDS Relief, January 2009 – December 2009.rce:
Quarterly Country Summary: Nigeria, 2008
 Has the program met its goal?
Percentage of new enrollees tested for HIV at each site, by
quarter
60%
% of new enrollees tested

50%
40%
for HIV

30% Site 1
20% Site 2
Site 3
10%
0%
Target
Quarter 1 Quarter 2 Quarter 3 Quarter 4

Data Source: Program records, AIDS Relief, January 2009 – December 2009..
quarterly Country Summary: Nigeria, 2008
Stacked bar chart
Represent components of whole & compare wholes
Number of Months Female and Male Patients Have Been
Enrolled in HIV Care, by Age Group

Females 4 10

0-14 years
15+ years
Males 3 6

0 5 10 15

Number of months patients have been enrolled in HIV care

Data source: AIDSRelief program records January 2009 - 20011
 Line graph
Displays trends over time
Number of Clinicians Working in Each Clinic During Years 1–4*

5
Number of clinicians

4
Clinic 1
3
Clinic 2
2 Clinic 3
1

0
Year 1 Year 2 Year 3 Year 4
*Includes doctors and nurses
Line graph
Number of Clinicians Working in Each Clinic During Years 1-4*

5
Number of clinicians

4
Clinic 1
3
Clinic 2
2 Clinic 3

0
Year 1 Year 2 Year 3 Year 4

Zambia Service Provision Assessment, 2007.

*Includes doctors and nurses
Pie chart
Contribution to the total = 100%

Percentage of All Patients Enrolled by Quarter

8%

10%

1st Qtr
2nd Qtr
3rd Qtr
23% 59% 4th Qtr

N=150
Interpreting data
 Adding meaning to information by making connections
and comparisons and exploring causes and consequences
Interpretation – relevance of finding
 Adding meaning to information by making connections
and comparisons and exploring causes and consequences
Interpretation – relevance of finding
 Does the indicator meet the target?
 How far from the target is it?
 How does it compare (to other time periods, other
facilities)?
 Are there any extreme highs and lows in the data?
Interpretation – possible causes?

• Supplement with expert opinion

• Others with knowledge of the program or target
population
Interpretation – consider other data
Use routine service data to clarify questions
• Calculate nurse-to-client ratio, review
commodities data against client load, etc.
Use other data sources
Interpretation – other data sources
 Situation analyses
 Demographic and health surveys
 Performance improvement data
Interpretation – conduct further research

 Data gap conduct further research

 Methodology depends on questions being asked and
resources available
Key messages
 Use the right graph for the right data
 Tables – can display a large amount of data
 Graphs/charts – visual, easier to detect patterns
 Label the components of your graphic
 Interpreting data adds meaning by making connections
and comparisons to program
 Service data are good at tracking progress & identifying
concerns – do not show causality
Planning for Analysis

Type of Type of
Data Formatting

Type of
Analysis
Type of Data & Formatting Technique
 Quantitative Data
 Must “quantify” the data
 Convert (“data reduce”) from collection format into numeric
database
 Qualitative Data
 Must process the data (type/enter/describe)
 Convert from audio/video to text
 Combination
 Process each element as appropriate
 Type of Data & Analysis
 Quantitative Data
 Counts, frequencies, tallies
 Statistical analyses (as appropriate)
 Qualitative Data
 Coding
 Patterns, themes, theory building
 Combination
 Process each element as appropriate
Quantifying Data
 Before we can do any kind of analysis, we need to
quantify our data

 “Quantification” is the process of converting data to a

numeric format
 Convert social science data into a “machine-readable” form, a
form that can be read & manipulated by computer programs
Quantifying Data

Some transformations are simple:

 Assign numeric representations to nominal or ordinal
variables:
 Turning male into “1” and female into “2”
 Assigning “3” to Very Interested, “2” to Somewhat
Interested, “1” to Not Interested
 Assign numeric values to continuous variables:
 Turning born in 1973 to “35”
 Number of children = “02”
Developing Code Categories
Some data are more challenging. Open-ended responses
must be coded.

 Two basic approaches:

 Begin with a coding scheme derived from the research purpose.
 Generate codes from the data.
Coding Quantitative Data
 Goal – reduce a wide variety of information to a more
limited set of variable attributes:
 “What is your occupation?”
 Use pre-established scheme: Professional, Managerial, Clerical, Semi-
skilled, etc.
 Create a scheme after reviewing the data
 Assign value to each category in the scheme: Professional = 1,
Managerial = 2, etc.
 Classify the response: “Secretary” is “clerical” and is coded as “3”
Coding Quantitative Data
 Points to remember:
 If the data are coded to maintain a good amount of detail, they
can always be combined (reduced) later
 However, if you start off with too little detail, you can’t get it
back
 If you’re using a survey / questionnaire, it’s a good idea to do
your coding on the form so that it can be entered properly (i.e.
create a “codebook”)
Codebook Construction
Purposes:
 Primary guide used in the coding process.
 Should note the value assigned to each variable attribute
(response)
 Guide for locating variables and interpreting codes in the
data file during analysis.
 If you’re doing your own input, this will also guide data
set construction
Hands-on Exercise 1
 Create a mini-codebook by coding the survey instrument
 Note column spaces / locations
 Note variable attribute values
 Pay attention to the box at the bottom, special instructions
Entering Data
 Optical scan sheets (usually ASCII output).
 Limits possible responses

 CATI system / On-line: entered while collected

 Data entry specialists enter the data into an SPSS data matrix,
Excel spreadsheet, or ASCII file.
 Typically, work off a coded questionnaire
Entering Data
 In Excel or Access, follow procedures from class:
 Format tables with proper variable columns
 Enter data for each case
 In SPSS
 Import an ASCII file and name variables/column headings
 Or, create variables/column headings & enter each case
Entering Data
 ASCII files are useful because they can be transformed or
used in almost all analysis programs
 Upload to SPSS, Excel, or use directly with SAS
Hands-on Exercise 2
 Complete the survey (fill-in your answers)
 Create a ‘dataset’
 Enter the data from your survey using either Notepad or
the Edit program from the Command prompt
Quantitative Analysis
 You should choose a level of analysis that is appropriate
for your research question

 You should choose the type of statistical analysis

appropriate for the variables you have
 Nominal/Categorical, Ordinal, or Continuous
Quantitative Levels of Analysis
 Univariate - simplest form,describe a case in terms of a
single variable.
 Bivariate - subgroup comparisons, describe a case in terms
of two variables simultaneously.
 Multivariate - analysis of two or more variables
simultaneously.
Univariate Analysis
Describing a case in terms of the distribution of attributes
that comprise it.
Example:
 Gender - number of women, number of men.

 You should always begin your analysis by running the

basic univariate frequencies and checking to be sure data
were entered properly
Univariate Analysis
 Frequency distributions

 Measures of central tendency

 Mean, Median, Mode
Presenting Univariate Data
Goals:
 Provide reader with the fullest degree of detail regarding
the data.
 Present data in a manageable from.
 Simple and straightforward
Subgroup Comparisons
 Describe subsets of cases, subjects or respondents.
Examples
 "Collapsing" response categories:
 Age categories, Open responses, etc.
 Handling "don't knows“
 Code separately, make missing if appropriate
Bivariate Analysis
 Describe a case in terms of two variables simultaneously.
 Example:
 Gender
 Attitudes toward equality for men and women
 How does a respondent’s gender affect his or her attitude toward
equality for men and women?
 Crosstabulations / Correlations
Constructing Bivariate Tables
 Divide cases into groups according to the attributes of
the independent variable.
 Describe each subgroup in terms of attributes of the
dependent variable.
 Read the table by comparing the independent variable
subgroups in terms of a given attribute of the dependent
variable.
 DV goes in the rows, IV goes in the columns
Bivariate Analysis
 Bivariate Tables / Crosstabs are appropriate for all types
of variables, but the proper inferential statistic will vary
by variable type

 Continuous variables are typically made into categorical

variables for this type of analysis
 Recode variables
 Example: Create “Age” (18-34, 35-50, 51-65, 66+)
Appropriate Types of Analysis
Bivariate Analysis: Correlations
 Bivariate correlation analysis is appropriate for continuous
variables (interval, ratio)
 Other types of variables are often recoded into ‘Dummy’
variables (value 0 or 1) for these purposes
 Example: Gender becomes two variables ‘Male’ (1=yes) &
‘Female’ (1=yes)
 Present in Correlation Matrix
Multivariate Analysis
 Analysis of more than two variables simultaneously.
 Can be used to understand the relationship between
multiple variables more fully.
 Most typical: Regression analysis
Multivariate Analysis
 Ordinal (technically inappropriate but it happens),
continuous, dummy variables

 Type of regression analysis will depend on the type of

variables
 OLS (continuous)
 Logistic (other types)
Sampling
 What is your population of interest?
 To whom do you want to generalize your results?
 All students (18 and over)
 Undergraduates only
 Greeks
 Athletes
 Other
 Can you sample the entire population?
Sampling
 A sample is “a smaller (but hopefully representative)
collection of units from a population used to
determine truths about that population” (Field, 2005)
 Why sample?
 Resources (time, money) and workload
 Gives results with known accuracy that can be calculated
mathematically
 The sampling frame is the list from which the
potential respondents are drawn
 Registrar’s office
 Class rosters
 Must assess sampling frame errors
Types of Samples
 Probability (Random) Samples 
 Simple random sample
 Systematic random sample
 Stratified random sample
 Proportionate
 Disproportionate
 Cluster sample
 Non-Probability Samples
 Convenience sample
 Purposive sample
 Quota
Sample Size
 Depends on expected response rate
 Average 85% for paper
 FINAL SAMPLE DESIRED / .85 = SAMPLE
 Average 25% for web
 FINAL SAMPLE DESIRED / .25 = SAMPLE

Size of Campus Final Desired N

<600 All students
600-2,999 600
3,000-9,999 700
10,000-19,999 800
20,000-29,000 900
≥30,000 1,000
Bias and Error
 Systematic Error or Bias: unknown or unacknowledged
error created during the design, measurement, sampling,
procedure, or choice of problem studied
 Error tends to go in one direction
 Examples: Selection, Recall, Social desirability
 Random
 Unrelated to true measures
 Example: Momentary fatigue
Levels of Measurement
 Nominal  Interval
 Gender  Body Mass Index (BMI)
 Male, Female
 Vaccinations
 Yes, No, Unsure
 Ordinal  Ratio
 Personal health status  Number of drinks
 Excellent, Very good, Good,  Number of sexual partners
Fair, Poor  Perception percentages
 Last 30 days  Blood alcohol concentration
 Never used, Not in last 30 days, (BAC)
1-2 days, 3-5 days, 6-9 days, 10-
19 days, 20-29 days, All 30 days