Graph slides


	Choose ‘cut off’ date for Data Analyses
		When was last contact with participant? Usually add at least 3 weeks to that date for the cut off. This will help to capture stragglers or late responders.
	Just because you chose a cut off date does not mean that people who send your questionnaires after that can not be added. but this needs to be decided before the mailings as part of the protocol-otherwise it is cheating!
		Quality monitoring for accuracy so that the data is reliable. If the checked data is 95% accurate then should do about 10% of your returned questionnaires. If more mistakes more Quality monitoring.
		Sometimes people will write in things cross things out-therefore must do Data cleaning-there is much less data cleaning when humans enter the data-they have already done some evaluation!
		Recoding is when you reconfigure (collapse, dichotomize, put into tertiles) data to get at a specific question. How many people aged 70 responding to the questionnaire have bone cancer. It is the easiest to collapse the under 70 together and the 70+ groups together. Recoding is meant to make the data more meaningful, but when ever you mess with data-you mess with data!!.