Parametric and non-parametric are major classification of statistical procedures. Non-parametric tests can be defined as statistical protocols which are not dependent on assumptions on the form of the probability distribution. On the other hand, parametric test protocols are reliant on assumptions about the shape of the population in questions, usually assuming normal distribution (Creswell, 2014). The assumptions in parametric procedures are drawn from the sample of the population. For instance, in comparing means between two independent groups for systolic blood pressure, the groups include treatment groups and a placebo. The parametric procedure that can be implemented in this analysis problem is two sample t-test and the non-parametric procedure is the Wilcoxon rank-sum test.
Parametric statistical tests make the assumptions that observations made on the data are independent, and are randomly selected from the population. Normal distribution is not expected in non-parametric tests. Due to their insensitivity to the sample size, non-parametric tests are only recommended for small samples (Creswell, 2014). In assessing nursing burnout, a multifaceted phenomenon in healthcare, the study will adopt a data analysis plan for assessing the effect of nurse-patient ratio on the burnout level. The population of the nurses will be the means of difference in means. In the non-parametric test, hypothesis will involve comparison of rank (level of burnout), frequencies, and medians. The histogram for the burnout levels may vary for the male and female nurses even when the nurse-patient ratio is the same. For instance, the distribution of males may be more symmetric with a mean and median that are almost similar. For the female group, the median may be quite different from the mean. Even though the median for the two groups are similar, a parametric test is unreasonable since assumption of normality cannot be made. A non-parametric test such as Wilcoxon rank sum will provide a significant difference (p-value) when comparing the two groups. A parametric test based on the central tendency measure, mean, will be misleading for the hypothesis while a non-parametric test will give significance differences between the groups without making assumptions on the normality of population distribution.
We can do it today.
- Creswell, J.W. (2014). Research design: Quantitative, qualitative and mixed methods approaches, 4th edition. Sage Publications, Inc.