Answer

Research Methods in Public Administration

            The purpose of this paper is to compare and contrast the classical experimental design and the quasi-experimental design in terms of control of internal and external validity. A quasi-experiment can generate evidence on causal relationship when randomized controlled trials are impossible (Bärnighausen, 2017). The validity of a study refers to how well the findings among research participants represent true findings among similar study (Patino & Ferreira, 2018). The validity of a study has two domains; internal and external validity. Internal validity is defined as the extent to which the observed findings represent the truth in a given population that is being studied. On the other hand, external validity is defined as the extent to which findings of a study are generalizable to other people in different settings or to other populations (Patino & Ferreira, 2018).

            Similar strategies are applied in classical experimental and quasi-experimental designs in addressing the issue of internal and external validity. To increase internal validity, researchers ensure careful planning, adequate recruitment strategies, quality control, and data collection, data analysis and sample size. External validity can be improved through use of broad inclusion criteria, which may lead selection of a group of study participants who closely resemble real-life study population (Patino & Ferreira, 2018).

            In experimental studies, internal validity can be improved via proper randomization strategies. On the other hand, external validity of either experimental or quasi-experimental study tend to be based on judgement rather than computed statistics (Andrade, 2018). According to Bärnighausen et al. (2017), quasi-experiments can generate high value causal evidence with higher external validity than classical experiments since they characteristically distort the context of an intervention than the experiments. Quasi-experiments generate with high degree of external validity use data from ‘entire’ populations across all available settings or data on all patients. Quasi-experiment data that is collected from routine data systems such as population census, clinical records data and the intervention under investigation. In contrast, randomized controlled experiments usually select patients and sites and interfere in the intervention processes. Randomized controlled experiments can result in externally invalid results due to selection bias. To address this, the sample should be representative of the population that is being studied in a randomized controlled experiment (Bärnighausen et al., 2017).

            Randomized controlled experiments achieve exogeneity via the researcher’s action to randomize individuals to control and treatment groups (Bärnighausen et al., 2017). Research participants are randomly assigned to either control or the treatment group. All potential research participants have an equal chance of being assigned to any group. Random assignment helps to neutralize factors other than the dependent and independent variable, which enables direct inferences of the cause and effect (The Regents of the University of Michigan, 2021). While in quasi-experiments, exogeneity is achieved through policy, nature of the study, and practice. Exogeneity of exposure implies that a given exposure is not influence by any variable that is associated with the outcome or the outcome interest. Exogeneity of exposure mean that the confounding and selection bias is controlled for, without explicit control or having to observe for any confounding factors in the analysis (Bärnighausen et al., 2017). Random selection can help minimize selection bias (Handley et al., 2018).

            In random assignment of research participants in an experiment, it is assumed that the independent variable is the only cause of the observed findings since the two groups (experimental and control group) may have not differed from one another at the beginning of the experiment (The Regents of the University of Michigan, 2021). In an experimental study, threat to internal validity may be observed if participants refuse to participate or drop out from the study. Dropping out from the study results in differential attrition. Better participant engagement and provision of background information about the study may prevent cases of participant drop outs (The Regents of the University of Michigan, 2021).

            Internal validity may be threatened by many factors including errors in selection of research participants in the study, and errors in measurement. Lack of internal validity means that the findings of a study deviate from the truth; thus, generalizations cannot be made. If results are not internally valid, then they will also be not externally valid. Lack of external validity means that the findings of a trial cannot apply to patients from a different population from the population under study (Patino & Ferreira, 2018).

            Systematic error can be a threat to internal validity of both experimental and quasi-experimental studies. Systematic error can occur via performance bias, selection bias, attrition bias, and detection bias (Andrade, 2018). Common threats to internal validity in quasi-experiments include selection bias, history bias, maturation bias, lack of blinding, variability in interactive effects, and different dropout (Handley et al., 2018). Internal validity can be improved through modification of the plan of analysis. Besides, any form of bias should be eliminated or reduced (Andrade, 2018). Non-blinded experimental studies may also pose threats to experimental studies, thus threatening internal validity. This can be resolved through use of double-blinding strategies (Bärnighausen et al., 2017). In addition, artificiality may threaten external validity in a randomized controlled trial due to use of processes and procedures that do not exist in the eventual real-life implementation of tested intervention. The use of inclusion and exclusion criteria can help address the problem of selection bias (Bärnighausen et al., 2017). In non-blinded randomized controlled experiments, internal validity in estimation of the effect size can be threatened, since trial participants may react to their assignment to either the control or the treatment group.  In contrast, quasi-experiments are often “blinded” and this helps in avoiding such threats (Bärnighausen et al., 2017).

            In conclusion, a balance between internal and external validity is important in a study. Randomization in experimental studies can be achieved through randomization of subjects into control and intervention group. In both experimental and quasi-experimental studies internal validity can be increased through careful planning, adequate recruitment strategies, adequate quality control, and adequate data collection, data analysis and sample size. Similarly, external validity can be increased through the use of broad inclusion criteria and use of a large sample size. If internal validity is invalid, then external validity is invalid.