DNP 830 Topic 5 DQ 1: Statistical Analysis Plan

Answer in the order the prompt gives, because each part constrains the next. Descriptive analysis comes first and is not a formality: it establishes who was in the sample, characterises the baseline, and shows whether the data meet the assumptions the inferential test depends on. Say which descriptives you will report and why — frequencies and percentages for categorical variables, means and standard deviations for normally distributed continuous ones, medians and interquartile ranges when they are skewed. That last choice is itself a finding about your data, and reporting a mean for a badly skewed variable is one of the commonest errors in these projects. Say which variables are categorical and which are continuous before naming any statistic, because that single classification decides most of what follows. Then state the sample size you expect, since it constrains the test as tightly as the design does and a doctoral project rarely has the numbers to ignore it.

Then the inferential test, justified from three things rather than named. The design decides the family: a single group measured before and after needs a paired test, two independent groups need an independent-samples test, and repeated measures across three or more points need a repeated-measures approach. The measurement level decides the variant, since a continuous outcome and a proportion do not take the same test. The sample size and distribution decide whether a parametric test is defensible or a non-parametric equivalent is safer, and a doctoral project sample is often small enough that this matters. Say all three and the alignment question is answered. Then close on the clinical question: a significant p-value says the result is unlikely under the null, not that the change was large enough to matter. Report an effect size or a mean difference in the outcome's own units. Say which you would report first in a manuscript, since ordering signals what you think the finding is.