Pollsters respond to those who criticize the size of their samples, and how can a small portion, of society, express millions of others, that the doctor should examine the entire person’s blood and not a sample of it! Despite the difference in context between social studies and medical studies, the issue of selecting samples is not as easy as some might think, and it does not deserve criticism from others!
There are several criteria that govern the sample size, such as the number of questions, the nature of the community, the respondents, the level of statistical analysis that we will use, the level of relationships and links and their complexity, and the degree of homogeneity or diversity of the community. The greater the homogeneity of the community, the smaller the sample size, and the greater the diversity and differences in it, the greater, in addition to other factors, such as the time and financial frame, the degree of precision we need, the difficulty of the questions and the potential for bias, and the relationship with demographic data (gender, age, etc.).
These criteria are studied in detail during the planning and preparation phase of survey studies, and some of them are presented in justifying the number of the sample, in addition to the ideal way of selecting it, and its design so that all members of the community are in one basket and have the same opportunity to be asked according to random selection.
Due to the accuracy of these standards, it has become common for there to be a statistician within the work team who follows up any study from its beginning, studies the population in preparation for drawing the sample, and discusses all these standards collectively with the researchers.
There are incorrect and widespread standards in this field, such as the statement of choosing 2% of the population as a sample! This is a percentage that may translate into very large numbers in communities numbering in the millions, and this is illogical and not common in the practices of any polling bodies.
On the other hand, there are equations suggested by some of them, such as for each simple question to have 30 items, meaning that 300 items are sufficient for a form consisting of ten simple, uncomplicated questions, but if the complexity increases, we may need ten times this number, and to go up from 300. To a thousand, and some may demand an increase to this number, and they respond that increasing the number after the thousand to 1,500 or 2,000 will lead to a decrease in the percentage of statistical change and its results. Some major bodies in measuring public opinion have tried drawing a sample from thousands of items, but they found that the financial cost and time wasted are not worth it, and some studies have found that increasing the sample size may lead to errors in data collection and increased bias, especially if some traditional methods are used.
If the frequencies are general and descriptive, the sample size can be negligible, especially if the available budget is reduced or the available time is limited. Also, increasing the sample size does not necessarily lead to a large number of responses, especially in sensitive surveys. In some of the electronic surveys that I supervised; the link was sent to 20,000 people. Randomly sent an SMS message, and only about 500 of them responded.
As one of the leaders in public opinion measurement (the former CEO of Gallup) said, achieving a perfect sample is a miracle if it’s done right, it’s miracle.
Dr. Samir Abu Rumman, General Supervisor of the World of Opinions Center
