MODULE 3 · LESSON 3.6
Estimated time: 30 minutes
Sometimes one dataset does not provide enough information to answer a question. A fictional health programme wants to understand service delivery in three communities. It has facility summary reports, community survey responses and NGO outreach records. Each source describes a different part of the programme. Comparing them may provide a more complete picture, but only if the information is compatible and the limitations are understood.
We will explore this topic in greater depth in Module 4. For now, the focus is on preparing datasets for comparison.
Service contacts (facility tallies) are not the same as unique individuals (catchment surveys). One person can appear several times in a tally. A survey may count each person once. Do not subtract one from the other and call the difference people missed.
Suppose one dataset covers January to March, while another covers April to June. You should not compare their totals as if they represent the same period. First, identify the period covered by each source.
Suppose one dataset covers three communities, while another covers an entire local government area. The figures may not be directly comparable. Confirm the geographical units and the populations represented.
Suppose one dataset counts the number of service contacts, while another estimates the number of unique people receiving the service. These are not the same measure. One person may have several service contacts. Before comparing the numbers, understand exactly what each source counts.
| Facility | Month | Patients |
|---|---|---|
| Central Clinic | January | 120 |
| Riverside Clinic | January | 95 |
| Facility_Name | Reporting_Period | Patient_Count |
|---|---|---|
| Central Clinic | January | 125 |
| Hilltop Clinic | January | 110 |
The columns have different names, but they may represent comparable concepts. Before combining them, verify that the facility identifiers match, the reporting periods are consistent, the patient-count definitions are equivalent, and the sources do not overlap in a way that would create duplicate counts.
Do not assume that similarly named columns necessarily measure the same thing.
| Source | Facility | Month | Reported attendance |
|---|---|---|---|
| Dataset A | Central Clinic | January | 120 |
| Dataset B | Central Clinic | January | 125 |
The difference is five recorded visits. We do not yet know why. Possible explanations include different extraction dates, a correction in one source, different counting rules, a transcription error, or different coverage. These are hypotheses to investigate, not conclusions.
Before comparing or combining datasets, check the reporting periods, geographical coverage, indicator definitions and data structure.