1 of 20

Data management: Preparing your data for analysis

Tigist W. (MD, MPH)

November, 2022

2 of 20

Content

Definition of data management
Data coding
Data cleaning
Data processing and organization

3 of 20

Data management: Definition

Collected data…………Analyzed (to answer our research question)

Coded
Cleaned and validated
Stored
Processed and organized

Data management software: Epi Info, SPSS, STATA, SAS…

4 of 20

Data coding

Raw data 🡪 Group (categorize) 🡪 Assign codes (0,1,2…)

Purpose

Statistical software read numerical codes rather than non-numerical information

5 of 20

Data coding cont…

Categorical variables

Three possible types

1. Already coded (Structured questions)

E.g., Gender: 1. Male, 2. Female

2. Partly coded (Semi-structured questions)

E.g., Drug discontinuation reasons: 1. Drug side effect, 2. Unavailability of Drug, 3. Cost of drug, 4. Other, Specify______________

2. Uncoded (Open ended questions)

E.g., Drug discontinuation reasons:________

Numeric variables

All are uncoded (Open ended questions)

E.g., Age (in years):_______

Possible types of questions

6 of 20

Creating codes

Partly open ended questions (other group responses)
Entirely open ended questions

Numerical values

E.g., Age (in years) ____ 🡪 10-19, 20-29, 30-39, …
How to create the categories?

Statistics (Sturges rule) Vs Science (clinically meaningful categories)

Drug discontinuation reasons	Frequency (%)
Drug side effect	120 (30%)
Unavailability of Drug	95 (23.8%)
Cost of drug	83 (20.8%)
Other
Religion	85 (21.3%)
Other reasons	17 (4.25%)
Total study population	400 (100%)

Data coding cont…

7 of 20

Data coding: Coding tips cont…

Give each data entry a unique ID

Create functional (analyzable) group

E.g., Which of the following symptoms do you have? 1. Fever, 2. Cough, 3. SOB, 4. Headache

Do you have fever: 1. No, 2. Yes
Do you have cough: 1. No, 2. Yes
Do you have SOB: 1. No, 2. Yes
Do you have Headache: 1. No, 2. Yes

0, 1 or 1, 2🡪 Dependent variable code: 0, 1

Mutually exclusive, exhaustive, and precisely defined categories

E.g., Age category

20-29, 30-39, 40-49 Vs 20-30, 30-40, 40-50

8 of 20

Data coding: Coding tips cont…

Consistent coding E.g.,
Fever: 1. No, 2. Yes and Cough: 1. No, 2. Yes

Fever: 1. No, 2. Yes and Cough: 1. Yes, 2. No

Hypertension: 1. No, 2. Yes,
BMI: 1. Normal, 2. Underweight, 3. Over weight 4. Other

Hypertension: 1. No, 2. Yes
BMI: 1. Underweight, 2. Normal, 3. Over weight

Giving the reference category the smallest code for all variables.

Missing value coding

- Give it a unique code

9 of 20

Practice

Data coding

https://forms.gle/qRjBGVL4qA2spgTT9

9

10 of 20

Data cleaning �(data cleansing/ wash/ scrubbing)

Detecting and managing errors

Has to be done carefully not to lose valid data.

11 of 20

Data cleaning cont…

12 of 20

Practice

Data cleaning

https://forms.gle/mcan91AaMcGivMc56

12

13 of 20

Data processing and organization (preliminary analysis)

Step 1: After data is coded and cleaned, run frequency for all variables to check the following;

Small frequencies

Responses with < 5%

Coding errors

Step 2: Then do Data transformation as needed

Creating new codes

To code partly/entirely open ended questions & numeric variables

Merging groups

To handle small frequencies

Recoding

To create new categories and correct wrong/inconsistent codes

14 of 20

Data processing and organization: Handling small frequencies

Handling small frequencies

What to do?

Transform the data🡪 Merging, recoding…
Do not consider for further analysis

Drug discontinuation reasons	Frequency (%)
Drug side effect	160 (40.0%)
Unavailability of Drug	15 (3.75%)
Cost of drug	17 (4.25%)
Religion	105 (26.25%)
Lack of family support	103 (25.75%)
Total study population	400 (100%)

15 of 20

Data processing and organization: Data transformation

Creating new codes

Coding open ended questions
Coding numerical values
Creating new codes from other response groups
Creating other group from already assigned categories

Drug discontinuation reasons	Frequency (%)
Drug side effect	120 (30%)
Unavailability of Drug	95 (23.8%)
Cost of drug	83 (20.8%)
Other
Religion	85 (21.3%)
Other reasons	17 (4.25%)
Total study population	400 (100%)

Drug discontinuation reasons	Frequency (%)
Drug side effect	160 (40.0%)
Unavailability of Drug	15 (3.75%)
Cost of drug	17 (4.25%)
Religion	105 (26.25%)
Lack of family support	103 (25.75%)
Total study population	400 (100%)

16 of 20

Data processing and organization: Data transformation cont…

Merging groups

In to a new labeled group
Adding already assigned response into already created other group.

Drug discontinuation reasons	Frequency (%)
Drug side effect	160 (40.0%)
Unavailability of Drug	15 (3.75%)
Cost of drug	17 (4.25%)
Religion	105 (26.25%)
Lack of family support	103 (25.75%)
Total study population	400 (100%)

Drug discontinuation reasons	Frequency (%)
Drug side effect	190 (47.5%)
Unavailability of Drug	169 (42.3%)
Cost of drug	12 (3.0%)
Other reasons summed up	29 (7.25%)
Total study population	400 (100%)