Longitudinal PMA Data Formats
Panel female data
Users can continue to access PMA samples as cross sectional surveys as before through our website, and they will not have a new format. On the samples selection screen, choose the Cross-sectional radio button and proceed with constructing your extract. Each observation will be a person at one point in time.
The online data dissemination system provides the longitudinal data in two formats: wide and long form.
Wide form merges variables from across time onto the same row of the panel woman, so that there is one record per woman. Each time-varying variable will have a number suffix (_1, _2, etc) to indicate which survey it came from. For example, INTFQMON_1 is the interview month of the baseline survey, and INTFQMON_2 is the interview month of the follow-up survey from one year later. The unique identifier for panel women is FQINSTID. FQINSTID is not available for household members who were not participants in the panel study.
If household members who are not panel women are included in the extract, their records will not be linked between survey rounds. Their information will only be available in the variables from the phase in which they were enumerated.
See our video on the longitudinal data for more details on how to access panel data from the online data dissemination system.
Long form includes multiple observations of the same woman on different rows but in the same extract. This can be useful for some types of longitudinal analysis, like survival analysis.
Service Delivery Point wide form data
PMA's sampling design for surveys conducted between 2013 and 2018 collect data from households and service delivery points from the same enumeration area (EAID, the primary sampling unit) for four rounds in a row before randomly selecting a new set of EAs.. Often, the same facilities were sampled in multiple rounds. The variable FACILITYID identifies the facility within rounds that were conducted in the same set of EAs.
Beginning with the redesign in 2018, PMA refocused on longitudinal data collection with purposeful repeated interviews of the same facilities over time.
When downloaded from IPUMS, the data file format for wide form service delivery points is similar to the wide form female panel data. Up to four rounds of data wer linked by FACILITYID. Variables include suffixes indicating Nth survey the observation comes from, not necessarily the survey round contained in either ROUND or PHASE. For example, for Burkina Faso 2014 to 2016b, YEAR_1 would be 2014, whereas for Burkina Faso 2017 to 2018, YEAR_1 would be 2017 (see the table below). The same set of EAs were sampled in Burkina Faso in 2014, 2015, and twice in 2016. Another set of EAs were surveyed in only 2017 and 2018 before a change in sampling design and the selected EAs in 2019.
|7627||Burkina Faso||2014||2015||2016||2016||Regional hospital||Regional hospital||Regional hospital||Regional hospital|
|7596||Burkina Faso||2014||2015||2016||2016||Surgery center||Surgery center||Surgery center||Surgery center|
|7982||Burkina Faso||2017||2018||.||.||Health center||Health center||.||.|
|7447||Burkina Faso||2017||2018||.||.||Private health center||Private health center||.||.|
|854161010||Burkina Faso||2020||2021||.||.||National hospital||National hospital||.||.|
|854170030||Burkina Faso||2020||2021||.||.||Private clinic||Private clinic||.||.|
Users should note that for a small number of cases (less than 4% of total cases), linked records of facilities had different EAID codes over time. Other details such as facility type or managing authority also differ over time.