Hi ,
Two healthy adults of the same height and weight won't necessarily have the same lung capacity.
That's because lung capacity doesn't just depend on body size. It also depends on factors such as age, gender and ethnicity.
If a doctor tested the lung capacity of a 60 year old Asian woman using a test calibrated
for 20 year old Caucasian men, the patient may be diagnosed with a respiratory condition, even if her lungs are perfectly fine.
That doesn't mean there's anything wrong with the test.
It
means the assumptions underlying the test (i.e. that it would be applied to 20 year old Caucasian men) weren't met.
Here's the thing...
No
model or test is 100% accurate under all circumstances.
Whenever you fit a machine learning model, you make implicit assumptions about how that model will be used through your choice of training data. If those conditions aren't met, the accuracy of the model will suffer.
Everyone is interested in knowing the accuracy of a model or test. But it's just as important to know the circumstances under which that accuracy measurement holds true.
Talk again soon,
Dr Genevieve Hayes.
p.s. I'm taking a break over the Christmas/New Year period, so this is my final email for 2022. Thank you for all of your support this year and keep an eye on your inbox for my first email of 2023 in mid-January.
p.p.s. Thanks to Dasun Premadasa for inspiring this post.
