Results from social science research based on observations are considered to be the most truthful. The results are not hypothesized, as scientists are not sure what to expect from behavioral observations, but a large amount of information can be obtained about one behavior of interest. Randomized studies, or studies that determine a group that receives experimental treatment, also assign a control group or a group that does not receive experimental intervention. Control group results act as comparisons by which the validity of experimental group results are judged.
Observations from social science research are hard to generalize to broader populations. This is especially the case in the event that the interaction between two people are under study. Trend and cohort studies are the social science research results that are most easy to generalize. Both study types collect information from observed study populations. They also each collect data over different periods of time making it easier for researchers to identify behavioral patterns. Data collection based on population allows for evidence based generalization to specific groups.
Covariance designs often use experimental research that subjects one study group to an experimental protocol and another to a placebo. This means that tests are issued to both groups before and after the studies. Test results give researchers correlation measures. Behavioral studies of cohorts are also excellent for making correlations. Cohort studies are more ethical in cases where the subjects of contention -- such as cigarette smoking -- are detrimental to study participants. Instead of introducing cigarettes to an experimental group, cohorts of people who already smoke are identified and their behaviors are followed over time.
Establishing cause and effect is another determination sought by social science research. A correlation identifies similar variables between sample populations, but proving cause and effect is much more difficult. Studies of trends and cohort populations provide raw data, but true cause and effect can be blurred by extraneous data that cannot be controlled by researchers. Behavioral research that uses covariance designs provide the most accurate platform for establishing cause and effect relationships. Tests administered to randomized experimental groups allow for cause and effect association while control groups provide comparison data.