My study on a simulation as an instructional intervention for neonatal kittens has undergone a massive change this week. In working through the literature for Chapter 2, I began to notice that my idea to explore feelings of safety in learning with a simulation rather than with a live neonatal kitten were moot. All literature pointed to these feelings of safety being valid at face value. I had planned to explore this through qualitative methods via participant surveys and turn this into measures of confidence in beginning with live neonatal kittens. Being unable to find any literature beyond face value statements on simulations promoting safety over working with live subjects (which in hind sight I should have completely expected), I began to re-work my study by going back to the Concept-Construct-Variable model.
- My Concept is remaining the same: neonatal kitten mortality rates.
- The Construct before was safety through training. Now, my Construct is: quality of care through training.
- The Variable then had to change, although my independent variable is the same (training with the simulation), the dependent variables would be different. Learners would be tested on their ability to:
- Sequence procedures correctly;
- Diagnose conditions; and
- Make tactical decisions.
This altered my research questions to focus still on neonatal kitten mortality rates, but also on those three measurable dependent variables in their capacity critical to neonatal kitten survival.
This alters my study from mixed-methods to quantitative as everything can now be performed through pre- and post-testing with data sampling of the mortality rates. I still desire to conduct a true-experimental study but that will depend on my sample size. At the very least, I will have randomized assignments to the simulation and control groups.
Random sampling, like random assignments, requires completely blind selection: not every third, not as they walk-up going to A or B, but completely randomized as they would by being picked blindly from a hat. Thankfully, there are free software programs with advanced algorithms to accomplish this such as the Research Randomizer. Using full randomization and algorithms can help secure reliability and validity of a research study design, which can help make it more statistically significant. Larger statistics can also help with validity: “As the sample size increases, and the standard deviation remains constant, the standard of error of the mean decreases” (Urdan, 2010, p. 59). Many studies use sample sizes that are just too small to return valid results. Other common threats to the internal validity of a study include: using personal judgment to select the sample which might misrepresent the population; creating testing instruments which can mislead participants; using testing instruments the participants don’t understand; or using instruments not previously proven reliable. Qualitative researchers must be extremely careful with their instruments for validity and reliability due to personal bias. Many use triangulation of their methods to have not only one alternate source of measurement, but a second.
References
Urdan, T. C. (2010). Statistics in plain English (3rd rd.). New York, NY: Routledge.