Surveys

This category runs from questionnaire design to a clean data file: choosing how many points your Likert scale needs, writing items for each construct so that each one measures a single thing, calculating sample size from the number of indicators and from the analysis you intend to run, distributing online, screening out invalid responses, and coding the result for import.

The posts are direct about the parts that go wrong. Response rates that come in well under what the sampling plan assumed. Reverse-worded items that confuse careful respondents more than they catch careless ones. Missing data, and the difference between dropping a case and imputing a value. Where a post gives a formula it names the source of that formula rather than presenting a number as settled.

If you would rather have an instrument distributed and cleaned than build that step by hand, some posts mention survify.net. That is our own product rather than a third-party recommendation, and every post that mentions it says so.

Once you have a usable file, the SPSS category picks up at reliability testing. The Research models category is the one to read first if you are still deciding which constructs the questionnaire has to measure at all.

Where to start. Read the post on Likert scales first, because the number of points decides how you word every item and how you code the responses later. Sample size next, and read it before you distribute rather than after, since too few cases is the one error no analysis will fix. The questionnaire design post covers item order, reverse-worded items, and screening questions. When the responses are in, the posts on response screening and on coding for SPSS take you to a file that is ready for Cronbach's alpha.