While the classical statistic is not anymore alone in the field of Data Science, it still plays a major role in situations, where no data is available about a certain problem. This is mainly the case in situations when one wants to evaluate how a new product is accepted by customers, what they think about a specific case, etc. In contrast to situations where the new core competencies of Data Science like Data Mining or even Big Data are applied, the number of cases is here much smaller. Usually these cases are not collected as a by product, but need to be actively acquired by means of an interview, questionnaire or experiment. With the smaller number of cases it is much harder to find effects and patterns in the data and it is an imperative to approach the problem from the very beginning in the right way.
Start in the right way
If a question wasn't asked during the interview, you cannot evaluate it's answer. While the sentence sounds stupidly simple, this fundamental (and trivial) truth needs to be adhered from the very beginning, but many times isn't. If one doesn't plan the data acquisition carefully in accordance with the target of the entire analysis, no technology will help to uncover the missing data. And of course there are many more pitfalls along the way, from introducing a selection bias by the means of acquiring the participants, from the way of asking questions and the form of the answer until confusing correlation and causality in the end report.
Our experts guide you through the entire process of a statistical project utilizing the benefits of lifting classical statistics onto a modern Data Science platform:
- Understanding the research goal
- Formulating hypotheses about the research goal
- Defining a suitable instrument to evaluate the hypotheses
- Test driving the instrument and optimize according to the test results
- Identifying suitable infrastructures and ways to acquire a sample with our instrument
- Evaluate the results