0
$\begingroup$

If we imagine an outdoor race with two obstacles:

  • If the participant fails an obstacle attempt they exit the race
  • Historical data shows that about 50% of participants will fail each obstacle
  • That means that about 25% will finish the race (.5 * .5)

100 people entered today's race with the following current status:

  1. 75 people have attempted obstacle 1 but only 40 succeeded
  2. 20 people have attempted obstacle 2 but only 8 succeeded (and have completed the race)

Per the numbers above:

  • 25 people are still waiting to get attempt obstacle 1
  • 20 people are still waiting to complete obstacle 2

What approach should we use to forecast the number of people who will complete the race?

update

Here is my thought process, starting with obstacle 1:

success=40 #succeeded
fails=35 #failed

#prior (50/50 prior)
alpha=10
beta=10

x = np.linspace(0,1,100)     
pd.DataFrame(stats.beta(alpha+success,beta+fails).pdf(x)).plot.line(
    figsize=(12,6),
    color='grey', 
)

enter image description here

The probable success rate therefore lies somewhere between 40% and 65% roughly, meaning that between 10 and 16 of the remaining 20 competitors will pass the first obstacle.

Is that right???

When calculating the same for obstacle 2, does obstacle 1 become a prior or is it completely unrelated???

Many thanks

$\endgroup$
  • $\begingroup$ Since you are given no historical information at all about people who are "still waiting to get through" various obstacles, why do you suppose any forecast is possible? $\endgroup$ – whuber Feb 17 at 20:05
  • $\begingroup$ thanks @whuber - based on my knowledge thus far, a Bayesian approach allows us to assign a 50/50 prior and we now have data on 75 competitors who have attempted the obstacle $\endgroup$ – DrBorrow Feb 17 at 20:09

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.