It depends whether you think that the community you are studying is the entire population relevant to your research question, or is just a sample of that population, from which you wish to make an inference about the actual population parameters.
Cross sectional study may require statistical test however KAP surveys normally should not require any statistics except mean/sd etc. If you rate the knowledge on a scale of let say 5 (Best=5...worst=1) then vector analysis may require.
Good question. The analysis will depend on what variables have been studies? what research questions have been risen? what are the level of measurements for the variales? what scales have been used?
Therefore, the descriptive statistics is a must.
The inferential statistics depends on the above points.