Because reliable salary information on what high-tech jobs pay in the Washington market is hard to come by, washingtonpost.com decided to offer tech professionals the opportunity to create their own online survey. The What Are You Worth? survey launched as part of the new WashTech section on April 22.

By September, we had more than 7,000 responses from the Washington area, so we set to work analyzing the results. First, we cleaned up the data. We took all entries from April 20 to Sept. 3, omitted records with internal inconsistencies, took out the non-full-timers for the time being, zeroed in on only the local responses, and set aside job titles with fewer than 30 responses.

Before you jump to the query page and delve in, please remember that this Web experiment is a self-selected survey, which means that rather than taking a random sample of tech professionals, we have been relying on you, the participants, to volunteer.

In one of our analyses of the database, we compared participants who submitted their information anonymously against participants who offered their e-mail addresses. Some survey analysts contend that "signed" responses are more reliable than anonymous ones. Our comparison showed that the submissions from respondents who gave their e-mail addresses are similar to the submissions from respondents who chose to remain anonymous. The averages differ, but one isn't consistently higher than the other. We took this result as a good sign.

We chose to offer the results in three formats: a median, a low-end salary, and a high-end salary. We ranked all of the salaries within a job title and took the salary at the middle of the ranking for the median. The low-end and high-end salaries correspond to the 5th and 95th percentiles respectively, resulting in more interesting statistics than the true minimum and maximum values, which tend to be exaggerated.

We are continuing to gather salary data, so please check back for future analysis and local tech salary trends. This kind of data only gets more interesting as time passes.