Hello Mathematics Geniuses,
I would like to know a way to construct a dynamic correlation matrix (ideally in Sql Server) so that I can compare each exam question (the likelihood of it being answered correctly) in relation to every other question.
As responses will be stored in a SQL Server table, it would be great if somehow the correlation can be done in SQL Server too, so as more people complete the exam the dataset on which the correlations are based can grow and become more reliable.
The starting data set will be very simple, along the lines of
Student_ID Question_ID Correct(Y/N) Timestamp
I am looking to find connections between pairs or larger combinations of questions, based only on the students ability to answer them correctly, thus avoiding superimposing assumptions about the relationships between different topics. The end result being that questions in exams conducted online can be ordered (dynamically) from easiest to hardest, based on responses made so far. It may also be used to skip questions that are likely to be correct and thus focus on other areas of greater challenge.
Apologies if that does not sound terribly clear and thanks for your time!