This word is often used to refer to data, but it is important to remember that "discrepancy" is NOT a substitute word for "difference"! A discrepancy specifically means that a conflict exists where one expects things to be the same.

Examples:

• You conduct a survey and compare your results to that of a similar survey conducted by the government. You notice there are significant (note: significant means "noteworthy", not necessarily "large") differences. You can say that there is a discrepancy between the two set of results.
• You perform an experiment to test a theory. You observe that your experiment produces results different from what the theory suggests. There exists a discrepancy between the theory and your observation.
• At a company, people holding the same job title are getting paid drastically different salaries. You suspect that some employees are being unfairly treated.. You can say that there are striking discrepancies between the salaries paid within a single job title at this company.
• A bar chart suggests that car accidents are more common than train accidents. You cannot say there is a discrepancy between the rate of car accidents and train accidents, because they are not related, and there is no reason why they should share similar statistics.
• A survey suggests a high-fibre diet is less likely to cause obesity than a high-fat diet. Again, there is no "discrepancy" between the health effects of high-fibre diets and high-fat diets because there is no expectation that they cause the same amount of weight gain.