Project Approach

All generalizations are false, including this one.

Mark Twain

Tackling new, intriguing questions every time – that’s what I love about my work. No two projects are alike; each client’s situation has its own specific issues and challenges. Which is why the mode of collaboration varies in each case.

Some basic elements are consistent, though: Every project starts with a discussion of problems to be solved; also, every project follows a clear structure. This structure can roughly be sketched out as follows, using the example of a rather comprehensive statistical modeling project:

1. Preparing and consolidating your data

To render meaningful results, an analytical project requires a clean and well-structured database. Bearing in mind the motto “as large as necessary – as small as possible,” it should include enough data to provide the information needed while allowing efficient analyses.

I review and structure your data, draft the first outlines of an analysis plan, and determine what other data will be needed to answer your specific questions. Also, I can take care of procuring any data available from official sources.

 

2. Reviewing data for quality and plausibility

Once the database is complete, I check the data for quality and plausibility, clarify inconsistencies, obtain any information that is missing, and make the necessary corrections. The result is a consistent basis for analysis.

3. Conducting descriptive and explorative analyses

Next, I analyze the data with regard to the set of questions initially discussed, using both descriptive techniques (condensing the information) and explorative approaches (processing the data to make patterns and relationships visible). The results of this process are new insights on the distribution of data, on the occurrence of “outliers”, and on meaningful correlations.

4. Formulating a model

The insights gained in step 3 then go into a statistical model. Using my statistical expertise, I work to reveal existing inter­dependencies between variables and assess their significance. To estimate model variables and perform statistical tests and forecasts, I use dedicated software such as Stata®, SPSS®, and the open-source statistics package R.

5. Refining the analytical approach and developing initial hypotheses

This step is concurrent with all previous steps: As the preparation and analysis of data keeps producing new insights, I iteratively adapt and refine my approach (in close coordination with you). Some initial hypotheses on the outcome will crystallize in the process.

6. Finalizing the model

Next, I systematically refine the statistical model I have built: Using special diagnostic tools, I check the validity of underlying assumptions, adapt the formulation if required, and enter new estimates until I come up with the final model. At this point, I will also confirm or modify my initial hypotheses and put them in precise terms.

7. Deriving statements and producing a report

In this last step, I use the statistical model to derive precise, actionable answers to your questions, which I summarize in a report.

Technically, I can start with any of the steps described above. However, as the overall duration of the project largely depends on the data quality and structure, I usually recommend starting with a data review.