Surveys and representativity of populations Flashcards
What is a survey?
A survey is a way in which one can identify a specific state off affairs. The main goal of a survey is to gather relevant information. In general, we tend to ask people some specific question to gather the desired information. Surveys can be conducted in several ways. Some examples are interviews (both physical and digital), mailed out queries, and web-based surveys.
What are are some of the pros and cons of digital surveys?
Digital surveys have become increasingly popular in the recent years since the offer several key advantages. Chief among which are cost, the instant access, wide audience, and flexible layout. There are however some downsides like privacy and anonymity (IP addresses are usually recorded to ensure that a single individual cannot reply more than once, this is at odds with GDPR). Low response rates is also a challenge.
Why was Literary Digest Magazine’s prediction of the U.S presidential election so far off the mark?
Literary Digest magazine got 2.4 million participants on a survey and still missed in their prediction by a huge margin. The main reason for this was that the poll targeted people registered in the phone book and as car owners. In 1936, only a small portion of the American population could afford such, and therefore, the poll only targeted people in the upper middle class which were more likely to vote republican over democrat. Consequently, there was a strong selection bias when finding participants.
How can researchers collect a representative sample?
Firstly, the size of the dataset is often touted as the key factor. However, there is no one size that guarantees representativity, it varies between research questions and their complexity. We only have some general rules for sampling: For small populations you should strive to ask as many as possible, however, for larger populations we generally find asking a few hundred people to be satisfactory. When we have our dataset, we should often strive to account for some prominent variables in the dataset. Some of the more common such variables are gender, age, where you live, and socioeconomic status.
What are the two main methods of sampling?
We have two main methods. The best are probability methods where the sampling is completely random. This significantly reduces the probability of biases affecting the sample. In general, probability sampling means that every member of a population has equal probability of participating in the survey.
Probability methods is seldom how surveys are conducted, however. Surveys are mainly conducted using non-probability sampling. The main sampling technique is convenience sampling where we simply take what we get. One sampling technique that tend to work better than probability methods, when we analyze a very specific phenomena that only affects a small portion of the population, is the snowball sampling technique. If we want to conduct some study on people who play video games to obsessive levels, we should utilize the snowball effect method. The main idea is that people who play a lot of video games tend to know others just like them. Therefore, by getting access to one such individual we may quickly get access to others in a similar situation. Thus, in such scenarios, the snowball effect method is better at getting a representative sample than probabilistic methods.
What are some key considerations researchers must make to ensure the validity of the findings from a survey?
We need to be careful when constructing the questions in the survey. Respondents tend to interpret the questionnaires intention, which can affect the response. The respondent may end up wishing to be kind to you or oppose you, or reason strategically before replying. Employees will probably answer somewhat strategically if the employer conducts a survey on the company’s security policy, happiness with the management, etc. In short, know that the way in which you ask your questions and the setting in which they are asked will affect the answers you get. People tend to be mindful of the answers as they will do whatever they can to avoid negative consequences on the back of saying something unpopular.
What are demand characteristics?
Demand characteristics are the archetypal roles participants of a survey fall into. There are four such roles:
a) The good participant role. Here the participant behaves in a manner such that your hypothesis will be confirmed. Here the participant does not want ruin the study. This is obviously undesirable as such behavior will affect the replies and thereby the results.
b) The bad participant role is the opposite of the former and is equally undesirable. Such participants may go as far as to try and destroy the credibility of the study.
c) The faithful participant role. Faithful participants follow the instructions presented by the researcher to the letter.
d) The apprehensive participant role is concerned with how the researcher may interpret the results that he alters his behavior in a socially desirable way.
What are some of the strategies to combat demand characteristics?
Hiding the research question has proved to be a useful technique to combat the effect of demand characteristics. This can be carried out by:
1. Deception to hide the research hypothesis
2. Post-experimental questionnaire where you ask the respondents what they believe to be the research hypothesis. Adjust your interpretation of the data based on the result of the questionnaire.
3. Minimize interpersonal contact between the researcher and the respondents. Have someone else conduct the survey for you.
4. Do not inform the person conducting the survey about the hypothesis of the research, such that the cannot tell the participants.
There are some ethical issues by holding back the research question.
