Synopsis Three: research questions and variables
Sometimes the toughest part of research is fashioning the question you wish to ask. It needs to be narrow and do-able, but not so narrow that it can't meet the "so-what?" test. It also needs to be answerable using research data. Oftentimes reading other work on the subject will help you to find a reasonable research question, and guide you toward ways to approach it.
Research questions may be close-ended (directional) or open-ended (non-directional). A close-ended question suggests a relationship between variables, such as "Is more violence shown by high-television-watching pre-schoolers than by low television-watching pre-schoolers?" An open-ended question does not suggest a relationship: "What is the correlation between television watched and violent behavior in pre-schoolers?"
Normally researchers begin by asking questions. Sometimes they proceed to hypotheses. A hypothesis is a statement suggesting a relationship between two (or more) variables. Usually it is advanced by researchers who are able to bring a strong grounding in theory to the topic, so that they are able to predict a relationship. Hypotheses are abbreviated H1, H2, etc., and may be two-tailed or one-tailed. The two-tailed H does not predict the nature of a relationship: "There is a relationship between the amount a television a pre-schooler watches and his or her level of violent behavior." A one-tailed H predicts a relationship: "The more television a pre-schooler watches, the more violent his or her behavior will be." These if/then statements, called "conditional syllogisms" in formal logic study, are common ways of fashioning hypotheses. Common research questions are formulated by asking "What is the relationship between ____ and ____?
A Hypothesis always includes a companion "null" hypothesis (H0) which suggests no relationship, such as "H0: There is no relationship between the amount of television a pre-schooler watches and his or her violent behavior." Interestingly, researchers do not try to "prove" the hypothesis to be true. Instead, they try to prove the null hypothesis to be false. If the null hypothesis is not supported by the data, the researcher can then claim the hypothesis is "accepted."
All research contains variables, that is, things that may vary. In fact, to examine these variables, however, they also must be quantifiable, "a concept that takes on two or more values" (Frey et al.). This does not mean you must examine variables using quantitative research methods. "Numerical" refers to variables which can be tested. For instance, if you ask the question, "How can politicians sell themselves to the public?" the answers can be nearly endless, and therefore untestable using research methods.
Independent variables cause change in dependent variables. Conversely, (at the risk of using the passive voice!) dependent variables can be changed by independent variables. This sets up a causal relationship. If we ask, "What is the relationship between the use of satellite television in developing countries and changes in literacy rates," clearly the usage of satellite television "causes" (or does not cause) a change in literacy rates. Usage is the independent variable; literacy rates is the dependent variable.
A "constant" is something which does not change. In the earlier example, "pre-schoolers" is a constant. Researchers try to establish as many constants as they can to isolate and measure variables. Sometimes a "confounding variable" creeps into the work: something not controlled by the researcher affects a dependent variable more than the selected independent variable.
After a research question or hypothesis is established, researchers choose a quantitative or qualitative method to gather data. This is a creative process: often more than one method can be used to answer a question, although some methods may produce more useful results than others. When researchers use more than one method to answer a question, it's called "triangulation."