Cause, effect, and causal relationship
In order to describe research designs, I first need to define what is cause, effect, and causal relationship (Shadish, Cook, & Campbell, 2002). A cause is a factor that occurs a certain condition. An effect is understood through a counterfactual reasoning (Shadish, et al., 2002). An experiment generally involves a treatment group that receive an intervention and a control group that does not. The counterfactual reasoning is to compare what happened between the two groups. If there is any difference in outcomes between the treatment group and control group, the effect is evident. Finally, a causal relationship is to prove that the cause and effect are ...view middle of the document...
, 2002). Random assignment allows us to assume that the only difference between the treatment group and control groups is the intervention. Any group differences occur by chance.
Another unique element is a pre-post design before the intervention (pretest) and after the intervention (posttest), in which intervention effects are expected in the treatment group. However, it does not necessarily include a pretest (Bachman & Schutt, 2013) because perhaps random assignment validates that the baseline conditions between the treatment group and the control group are the same. In regard to random assignment, there are some controversial arguments. For example, Sampson (2010: 490) states, “Randomization is not the scientific panacea” in eliminating confounding variables. I will discuss this issue in the internal validity section.
b. Cross-sectional designs
The purpose of using cross-sectional designs is to identify the number of cases of a variable within a given population(Mann, 2003). Examining the prevalence of arrest of a certain population would be an example. Cross-sectional designs are distinguished from experimental designs and other non-experimental designs for three aspects. First, cross-sectional designs are non-experimental designs. Thus, random assignment is not carried out. Second, they require neither an intervention nor a control group. For this reason, no intervention outcomes are expected. In cross-sectional designs, there are observed outcomes, but these are not manipulated by researchers or affected by an intervention.
Third, cross-sectional data are collected only once in time. This often makes it hard to assess causal inference. A common example is found in research on broken windows theory. Perkins, Meeks, and Tylor’s (1992) cross-sectional study on physical environments and fear of crime concluded that community physical incivilities increased perceived crime problems. However, since their data were collected at one point in time, it is unclear whether physical incivilities affected perceived crime problems or perceived crime problems affected physical incivilities through community indifference, or their relationships are reciprocal. Despite such weak causal inference, cross-sectional designs provide several benefits in terms of cost-effectiveness and efficiency for gathering multiple outcomes and large samples in a short time.
c. Longitudinal designs
Longitudinal designs are also sort of non-experimental designs. The purposes of longitudinal designs are (1) describing patterns of change and (2) identifying effect directions (Menard, 2002). Since they are non-experimental designs, random assignment and intervention effects are not considered, and there is no control group. It differs from cross-sectional designs because longitudinal data are collected at more than one point in time. It is also distinguished from time-series designs in that time-series deigns generally involve an intervention, whereas longitudinal designs...