companies in a certain geographic region, revenue bracket, etc.) until you eventually hit the jackpot and find that one of the hypotheses (i.e. You develop a hypothesis for each one (i.e. You then start dredging through all of your data - because you’re a modern data-savvy person, you’re storing hundreds of attributes about your customers. customers with more than 20 employees) and the test shows no statistical significance. You test one variable or customer characteristic (i.e. Let’s say you want to see if some subset of your customers is more likely to upgrade than others. What Data Dredging could look like in your business To avoid this, it’s now becoming standard practice to register clinical trials, stating in advance what the primary endpoint measure is. That’s why so many results published in scientific journals have subsequently been proven to be wrong. Historically, this has been a problem with clinical trials where researchers have ‘data-dredged’ their results and switched what they were testing for. You’ll start testing hypotheses around things like hair color or animal affiliations until you find the data point you wanted to find - that unicyclists are more likely to win the lottery, for example.
By looking at a huge pool of data such as lottery winners without defining what you’re looking for, you’ll draw a whole lot of data points. Tests for statistical significance only work if you’ve defined your hypothesis up front.
Data dredging is the failure to acknowledge that the correlation was, in fact, the result of chance. Slice your data in enough different ways and you’ll observe some correlations purely as a result of chance. It’s searching deep for answers in data regardless of whether or not there is really something to be found. In simpler terms, it’s beginning to analyze data without saying what point you’re trying to prove and whether or not that point is actually valid. We’ve put together a series of quick lessons to help you spot fallacies in data or call out dodgy visualizations so you can use data with confidence and make better decisions both in work and in life.ĭata dredging is the use of data mining to uncover patterns in data that can be presented as statistically significant without first devising a specific hypothesis as to the underlying causality. At Geckoboard, we’re on a mission to help people use data more simply and effectively.
0 kommentar(er)
