Research Design Types and Threats to Validity
A non-experimental research design is not appropriate for addressing threats to internal validity because it depends on a non-controlled environment and does not factor treatments to the data being researched. Typically, the predictor variables cannot be manipulated by the researcher, and the results of the study are obtained through observations and interpretations leading to certain conclusions. Threats to internal validity that happen in this case include ambiguity of interpretations that are not controlled, problems with the instrument, and selection bias that cannot be decreased, but rather are significant problems with the research design.
Besides, control groups cannot be used in the research study to assure the researcher of internal and external validity. However, Trochim notes that threats to construct and conclusion validity can be reduced by using or relying on assumptions and making external generalizations on the results. However, because quasi-experimental research design relies on one group pre-posttest design makes it is possible to minimize the threats to internal and external validity by using appropriate instrumentation for data analysis. On the other hand, quasi-experimental research design appropriately reduces bias and the use of confounding variables to make the research valid. This can be done by increasing content to make a substantial generalization that can be generalized across the population of interest. This reduces heterogeneity among the variables used in the study. Conclusion validity leads to the increase in the validity of the results that have been generated.
The rationale is that the threats to conclusion validity, which includes low statistical power, problems with units of study, and unreliability of the measures in use can be reduced significantly. True experimental research design relies on randomization, treatments, and measurements or observations. According to Trochim, the threats to this kind of research design can be addressed by removing the measurement errors using the variance of the true scores against the variance of the measured scores.