“Fabrication” means making up data and publishing studies that were never actually done. “Manipulation” means changing or omitting data. But you must not make up or falsify data or results. Never change data or results to try to support your argument.
If you decide to omit data so that the results of your analysis are not skewed, you need to explain why and how. For example, perhaps one of the participants in your study was lost to follow-up (they stopped participating). You may then wish to not include this participant’s previous data in your analyses. That is fine, but you need to explain the situation in your manuscript. Maintain transparency at all times.
Additionally, you must not change or manipulate data in graphs and images. This includes using computer programs to change parts of photographic images. Changing data, even slightly, is unethical. Your work will be discredited if these changes are discovered. Journal editors will consider retracting your work if they believe that data have been fabricated or manipulated. You must never change or falsely present data to improve your chances of publication. Just read the news headlines to see the disasters that may come. Even top researchers have destroyed their careers by altering their data.
As an author, you are responsible for all the published content of your paper. This includes your figures. Check that all your figures are accurate before submitting your paper. Journal editors or peer reviewers may ask to see your raw data if they suspect that any figures have been fabricated or manipulated.