This week I was part of the group that went over the methodology of Willwald’s statistical study and it occurred to me that it might be useful to go over some statistical definitions used. These kinds of studies are common in development work and I think it would be useful to talk about for future reference.
The goal of the study was to use empirical information to test certain hypothesis about gender. This is a very common task, the goal of which should not necessarily be to prove your hypothesis correct but to find the most accurate information. The two types of hypothesis are null hypothesis and alternate hypothesis. The first is the hypothesis that assumes the results from a sample observation occur purely from chance, while the second is the hypothesis that there is another outside factor influencing the sample. One useful example I found stated that if one were to determine if a coin was fair, the “null hypothesis might be that half the flips would result in Heads and half in Tails. The alternative hypothesis might be that the number of Heads and Tail would be very different.” From there one looks at the data and determine if it is a statistically significant difference from what you would expect: 50 Tail and 50 Heads.
In this case the study also uses simple random sampling to select the member of each house to be interviewed. This method means very much like it sounds: each person in the sample is assigned a number (randomly, typically though a computer program that generates numbers), which has an equal probability of being selected. This method is often used when the person collecting data does not know much about their population. However, it is commonly used because it is easy to implement and analyze.
It is also important to keep in mind that statistical data can be easily manipulated to make it work in your favor or facts may be left out. For example, in this study they mention that in one case 80% of the participants selected in Burikina Faso were male, which contradicts the expected percent of male according to the census data of 1:1. In this case they article brings up the discrepancy, but others may not be so honest. Statistical analysis is a very valuable tool, but always remember to take the interpretations of the data with the grain of salt.