Questions / Quant research, Research decisions
Use a paired comparison
On the same 25 independent test dates, model A minus model B has mean loss difference −0.4 and sample standard deviation 1.0. What is the estimated standard error of the mean paired difference?
Answer it in the app It's free. You get hints, a second try if you miss, and a follow-up question.
Show the answer and worked solution
Answer: 0.2 loss units (1/5)
- Each independent sampling unit is a date, with one paired A-minus-B loss difference.
- The standard deviation of these differences is already given as 1.0. Divide by √25=5.
- The estimated standard error is 0.2 loss units. Do not add a second variance for the paired model.
Interview shortcut. Paired SE = sd of differences/√n = 1/5.
Takeaway. Use the distribution of within-pair differences for a paired comparison.
In the app this has a follow-up where one thing about the setup changes. Try the follow-up.