Framing research problems: defining, designing, and data collection

Framing the Research Problem: Basic Steps

The specific steps undertaken when framing a research problem for a study will vary with the type of discipline, subject area of research, and the level of accuracy demanded of the research. For example, a small exploratory study designed to see if there is a market for a new fitness studio in a suburban area will demand a different level of scrupulosity than a statistical study designed to see if a new drug has dangerous side effects within certain demographic populations. However, broadly speaking, the steps of the research process are as follows (Marketing research, 2009, Quick MBA):

Define the problem

The problem must be framed in a clear question format, and the data the research is attempting to accumulate must provide a reasonable answer to that question. For example 'is there a statistically significant correlation between hours of television watched and a child's BMI (Body Mass Index)' or 'what characteristics do mothers say influence their breakfast cereal choice when shopping for the family' are both examples of research-based questions.

Most research is framed as a null hypothesis: in other words, the research statement is the opposite of what the researcher actually wants to prove. In the case of a study regarding television watching and childhood obesity, the null hypothesis might state that there is no correlation between hours of television the child watches and the likelihood that the child's BMI will be in the overweight or obese range. The null hypothesis often states conventional wisdom or the status of the control group.

Step 3: Identify data types and sources: Sample size and collection methods

Data must be drawn from a representative population with a wide variety of norms that could influence the results. For example, in studying the effects of fast food restaurants near schools, the study population of children with many fast food restaurants near their school should be balanced in terms of BMI, gender, socio-economic status and other factors that could influence the results. So must the control group. Of course, some research studies might wish to study intervention techniques in the context of specific population groups, like the effects of a diabetes management program upon a low-income population with a quite limited population base, demographically.

If an ideally balanced sampling is not possible to achieve, nonparametric statistics must be used to allow for an uneven breakdown of demographic data (Rumsey 2007, p.262). Chi-square tests create a balanced group from a large population sample, while the Fisher exact test allows the researcher to balance data of from small samples (Nonparametric statistics, 2010, Stat Soft). Nonparametric methods are useful when the researcher knows little of the parameters of the variable of interest in the population -- for example, if the socioeconomic status of the students might not be able to be known conclusively, given the wide variation that might exist within school districts (Nonparametric statistics, 2010, Stat Soft).

Step 4: Collect the data, analyze and interpret the data, prepare the research report

The raw data of the study must be formatted to eliminate errors and then tabulated. "Simple tabulations count the occurrences of each variable independently of the other variables. Cross tabulations, also known as contingency tables or cross tabs, treats two or more variables simultaneously" (Nonparametric statistics, 2010, Stat Soft). Interpretation of the data will be shaped by the type of study -- for example, to compare two independent populations which are fairly equal, a signed rank test or t-test will be performed while the Wilcox signed-ranks test is more commonly used with irregular population samplings (Rumsey 2007, p. 295). The final step is to determine if the results are statistically significant enough to disprove the null hypothesis and prove the hypothesis of the researcher.