6 Analysis

The right analysis follows from your design and the kind of data you have. Choose it before data collection, not after.

Before you move on

Your analysis matches your design: paired or independent, two or more time points, or data tracked over time.
You've looked at your data in charts before running any test.
You report effect sizes with confidence intervals, not only p-values.
For quality improvement data tracked over time, you use a run chart or control chart.
Any change from your planned analysis is explained and approved.

Choosing an analysis after seeing the results invites choices that make results look better than they are. Decide in advance, record the plan, and if you need to change it, say why.

Start every analysis by looking at your data. A simple chart often shows problems, such as data entry errors or extreme values, that no test will point out for you.

Guides for this stage

Building and reading a run chart
Plot comparable observations over time, keep a clear reference median, and recognize patterns without overstating causation.
Choosing a statistical approach for your DNP project
Choose an approach by starting with the question, observation unit, comparison structure, and kind of outcome.
Analyzing Likert data: Items vs scale scores
Distinguish an individual ordered response from a supported multi-item score before selecting summaries or statistical tests.
Paired t-test or Wilcoxon? Analyzing pre-post data
Understand what paired tests examine and why the distribution of individual changes matters more than a simple sample-size rule.