Eduard Szöcs

Data in Environmental Science and Eco(toxico-)logy

Statistics: It works, bitches!

One month ago R user Roman Luštrik asked on the Stackoverflow R chat for predictions when the R-tag will reach 100,000 questions.

I accepted the challenge, grabbed some data from Stackoverflow and build a very simplistic GAM (see here for details). My model predicted that on 22nd July 2015 the 100,000th question will be asked. A few days later I refitted the model with updated data and the prediction was 21st July 2015.

Roman’s prediction using ETS (Error, Trend, Seasonal) ExponenTialSmoothing was few days later, on 26th July.

The truth

After waiting a month we knew the date when the 100,000th question was asked:

Today, the 22nd July 2015.

This is exactly what my model predicted and I’m quite happy about it! In statistics one can always make predictions, but the best is when they are accurate and become true.

Although, the modelling was quick & dirty and it might have been worth a look at the uncertainty of the prediction: It works, bitches!

Written on July 22, 2015