Method 1. Random Sample

If you want your survey to be representative you need to select your sample with great care and in  accordance with “probability”.

There are different ways of selecting your sample:

• You simply ask all of the women in your choosen group, e.g. in a village or all students of your university, all doctors in your city etc.

• You select a sample by rough calculation taking into account age, education, ethnicity and other relevant factors.

All the women in the group

You have selected a village, a quarter, or a somehow defined group of women, for example the employees of a company. And you are going to ask them all. But even when conducting a survey in a small village not all women may be home (if only a few are not in, it doesn’t matter, but if half of them are out working, then this is not “all”).

However, this method only makes sense if the group you choose is somehow representative or meaningful to the purpose. If you happen to have a list of all businesswomen in your city and decide that this should be your sample you will get problems when interpreting the results. Businesswomen are not representative of society or a segment of society and they are not meaningful to anything related to FGM.

Representative of a segment of society could be students of a university if they come from different religious, ethnic and geographical backgrounds. They are not representative in terms of education and probably in terms of class background. But this is an obvious feature, people who read your study will immediately understand this. “50% of students have undergone FGM” makes sense. “50% or businesswomen have undergone FGM” will leave the reader with a big question mark: What does this tell us?

Meaningful are groups related to health – and in some cases religion. If you find that 50% of female doctors with daughters have their daughters cut or plan to do so, then this tells you a lot. You can surely conclude that the rate is higher in society if those who should know best about the harms of FGM endorse the practice.

Rough Calculation

Where are you going to find a good target group? In a village, town or city quarter? There are two things to think of:
You need to be aware that the survey area affects the outcome. If you choose a suburbian area you might get different results than in a remote village. It’s good to pick a number of places – rural and urban, rich and poor.

What is feasible? Does your time and money budget allow you to go to remote villages? Can you knock on doors in your preferred city quarters? When will people be home?

• If you are going from house to house, make a plan before you start.

• Take a map. Divide your survey area in several parts. Compile a list of the places you want to visit. At each place, a given number of interviews (depending on the population size) will be conducted.

• You will go to each place, then look for nearby residential areas to conduct the required number of interviews. Go from door to door, visit more than one street. If you select, then try to select the houses as randomly as possible (every 3., 4. …) . To keep it random, only question one woman per family, then go to the next house – or you will have to question two in every house.

Which ethnic or religious groups?

Be aware that there may be different social groups within the defined territory. Will you be able to give all of them appropriate representation in your survey?

For example if you have about 25% Turkmen in your town, you could include 25 women from the Turkmen community. Then you have no difficulties with calculations in the end (if your sample size is 100). While conducting the survey you may find it difficult though to find exact numbers which correspond with the percentage in society. In this case, you should make sure that a group is not too small. If you only interview two Turkmen, their answers tell you nothing about Turkmen in general: One of them could be incidentally the only Turkmen in town who has undergone FGM. If you would conclude that 50% of Turkmen have undergone FGM, this would be far from reality. If you interview 20 and 10 of them have undergone FGM, you are much more safe to say 50% of Turkmen in this town have undergone FGM (even though with such small numbers there may still be large errors: It could be in reality 40% or 60%, but at least you are not completely off).

• To get meaningful data each group should consist of at least 30 people.

Probably you will not be able to conduct a survey which gives you 100% exact figures in a statistical sense like a “big” survey with access to money, manpower and population data. So we are just eager to roughly represent groups in our survey that are obviously present in society. We should reflect the patterns of religious and/or ethnic minorities and their relation – again roughly – to other groups.

• If there is no official or reliable data about the population (for example: Check the CIA World Factbook for information on the age structure in your country), then you have to carefully guess it for the places you want to do the research.

Often, and that will be decisive for the outcome of your survey, it is a clear distinguishable (ethnic, religious….) group, that practices FGM, while other parts of society do not. If you know this for sure already, than you have a clear-cut hypothesis and you can concentrate on that one single group. But be aware that other groups may have adopted the practice even if it wasn’t their tradition before. Unless you are absolutely certain, it could still be useful to test your hypothesis by a small control sample (e.g. asking 20 women of other groups). If you just have some assumptions you should include all groups

• If there are more than 3 relevant ethnic/religious groups in your town, this system is not feasible because it is getting to complex and your samples for each group are getting too small. Leave it and focus instead on a different indicator like geography. Or you can choose the two or three biggest groups or the two or three groups where you suspect FGM is most prevalent.

• Now, You can either try to have each group represented according to the ethnic structure of the population. Or you divide your sample in three equally large groups each with about 30 interviewees.

Which age?

A survey claiming some degree of representativeness should try to represent the age structure of the population. Again you need groups: e.g. 18-30; 31-45; 46-60; above 60. The groups don’t have to have the exact same age span. It is more important that decisive changes for women’s role in society are reflected. For example: If at a certain point in your country’s history basic education for all girls was introduced, then those who were born before should be in a separate group.

You can try to have each age group represented according to the age structure of the population.
For example if you know that persons below 30 make up 70% of the total population, you could have 70 women of this group in your sample group and make calculation of the overall FGM prevalence easy in the end.

You can also try to have each group represented equally and thereby find out more about small age groups, e.g. the grandmother generation.

• Grandmothers: Reflect that they are often the driving and decisive force in the practice of FGM.

• Mothers: With younger mothers, you have information access to the next generation of girls, that will potentially undergo FGM.

• We recommend to interview women age 16 and older. For the FGM status of younger girls you will have to ask the mothers (this question to the mother is included in our basic questionnaire)

University Students

Instead of choosing a city quarter or village you may also choose university students as your target group. If your results are 50% of students in XYZ university have undergone FGM, this is just as meaningful as the result 50% of women in XYZ district have. Both results do not tell you the prevalence of FGM in the whole of society but they indicate a reliable trend.

The tricky part about surveying students is that you can’t go door to door. You can probably just meet them on campus – and if you simply interview the first 100 you bump into, this would be a convenient sample.

However, you can also apply the methods of the random survey. In this case, you should make sure you question women of different fields of study and group them like you would group ethnic groups. Then you need to develop a method how you pick them randomly. You could interview a certain number when they come out of class: e.g. pick 20 classes and interview five from each. If it is a small university you may think of a way to obtain all or most email addresses.

Next

• Think of another method? Read about convenient samples.
• If you are thinking of using the internet to contact people read more about online surveys.
• Or continue with: How to use the basic questionnaire.