The original design estimated the number of interviews (sample size) that would be needed to be a representative of the larger population of children and parents. These samples sizes were then divided by 10, the average number of people in a FGDs, and the research project determined the number of FGDs that would be needed to obtain prevalence rates would be a total of 980 FGDs. Yes, 980 FGDs among children and parents on violence against children.
In addition, the research protocols stated that the FGDs would:
- recruit participants randomly
- cover a wide range of issues
- include confidentiality
- not attempt to get individual accounts of violence
- would solicit consensus on type and contexts of violence
- data analysis would be completed in a short period of time
- The sampling was based on individuals not groups. The sampling parameters (confidence interval, margin-of-error, etc.) would not be applicable once you took individuals and formed them into groups.
- FGDs findings can not provide prevalence rates nor be generalized of a larger population.
- Random sampling does not apply to FGDs; instead FGDs use either purposive or convenience samples. FGD participants are selected because of some common characteristic(s) not randomly.
- Generally, FGDs should focus on a few issues with sufficient time to "dig deeper" into these few issues rather than discussing a wide range of issues lightly.
- FGDs are not good for discussing sensitive issues such as child violence or exploitation, which is better handled in private interviews or in-depth interviews.
- FGDs cannot ensure confidentiality of what is discussed. FGDs organizers and the moderator cannot control what participants may tell others what was discussed in the FGDs afterwards.
- FGDs should solicit as many diverse opinions and views and NOT attempt to impose consensus.
- Data generated by FGDs are not cheap and easy to enter, analyze and interpret. Despite what may project staff think and/or believe, qualitative data entry and analysis must be systematic and rigorous and is as challenging to analyze as quantitative data.
- A good FGD requires an experience moderator and consistency across FGDs, which would be almost next to impossible with 980 FGDs.
Anyone interested in the best practices of using FGDs should read, International Focus Group Research: a hanbook for the health and social sciences, by Monique Hennink.