Statisticians – structural engineers not decorators

What role does a statistician play in a project that requires obtaining and assessing quantitative evidence to address a research or policy question?

I find that the work required in building a house can be a useful analogy for such a project . In this analogy the statistician is the painter/decorator*.

Decorators make a house look pretty

My understanding is that the painter/decorator is typically brought in after:  the plans for the house have been designed; the materials to build the house with have been chosen; the foundations, walls, roof, doors and windows are in place; the plumbing and wiring has been done and the walls have been plastered.

The decorator’s task is then to finish the house off and make it look nice. Sometimes their role is to cover over the cracks or problems with wallpaper or damp-proof paint. There is nothing that can be done about the structure. Yes, they can make something look smaller or larger depending on the choice of paint colour but fundamentally the decorator has to do the best they can with what they have.

To my mind this is equivalent to the statistician being brought in towards the end of the project, once: the project objectives have been decided; the study design and methods for data collection chosen; the data collected (people surveyed, plants measured, experiments run etc) and the data put into electronic format. The statistician is then required to do “something” with the data. In an extreme case the statistician is just brought in to calculate the p-values to “decorate” the story that the data will tell – which has already been determined.

Of course, in many cases bringing a statistician in at the end of a project is all that is possible.  And, a statistician can be very useful at this stage. They can: find the story in the data; produce informative presentations (graphical or otherwise) to tell that story; and identify caveats in the story telling. Or less prosaically they assess the evidence to address the question of interest using the information available (ie data analysis); produce informative presentations to present and interpret that evidence clearly; and identify any caveats in the evidence and conclusions that are drawn.

It is just that sometimes it is such a shame that the statistician wasn’t able to be there earlier.

Structural engineers make sure the house has a firm foundation

If we return to the house building I think it can be useful to think of a statistician as a structural engineer.  From the national careers service a structural engineer is…

… a key part of the design and construction team, working alongside civil engineers, architects and other construction professionals… (working to)… create all kinds of structures from houses, theatres, sports stadia and hospitals to bridges, oil rigs and space satellites….

Similarly a statistician can play a key role helping to design and create the project. For example, they can:

• work with others to define the objectives of the project.
• help choose the best possible data collection methods to meet the project objectives so that time and money are not wasted on inappropriate materials, experiments, survey questions or other data collection activities.
• look ahead to help ensure that the data that will be collected has the capability to provide a clear and simple answer to the original question or at least provides the information to disentangle a complex state of affairs.
• raise awareness of the potential limitations of the project may be and ensure expectations are realistic.

Of course they also are then well placed to assess the evidence, interpret and present the results and identify caveats. If they have done their job properly at the planning stage this assessment and presentation of evidence should be more straightforward and bring up fewer surprises than a situation where they were not involved in the planning.

So including a statistician on the project team from the start can help you ensure that the foundations of your project (house) are sound and give you the best possible chance of meeting your objectives and presenting clear and interpretable results (i.e having fantastic decoration)!

What do you think?

*I know this is a bit of a caricature but I think it does, at some level, describe a view that some people have of statisticians. Certainly it is something I have experienced and although many people do have a very different view I think it can be useful to spell it out. I mean no offence to anyone who is a painter, decorator or structural engineer!