Ascertainment bias...i just don't get it!!

[Northerncardinal]

Can anyone explain Ascertainment bias? I've looked it up but getting different explanations. I just don't get it.

thanks so much!

---

Members don't see this ad.

 

[dempty]

Can anyone explain Ascertainment bias? I've looked it up but getting different explanations. I just don't get it.

thanks so much!

Ascertainment bias tends (in clinical trials) to refer to biased outcome measurement when the person doing the measuring (like the physician) has knowledge of the treatment a patient received. Imagine a study in dermatology for a new ointment to treat a skin rash vs a placebo. If the dermatologist assessing the improvement knows who received what treatment, it's possible that the placebo group gets lower improvement ratings and the new ointment group gets higher improvement ratings than they otherwise would if the outcome assessor had no knowledge of who received what treatment. (This is one of a few ways this kind of bias could occur.) Long story short, outcomes are systematically shifted from the true values, on average, because someone had knowledge of the treatment. Be careful, though as the definition does depend on context (in some cases being as simple as general sampling bias).

 

[Northerncardinal]

Ascertainment bias tends (in clinical trials) to refer to biased outcome measurement when the person doing the measuring (like the physician) has knowledge of the treatment a patient received. Imagine a study in dermatology for a new ointment to treat a skin rash vs a placebo. If the dermatologist assessing the improvement knows who received what treatment, it's possible that the placebo group gets lower improvement ratings and the new ointment group gets higher improvement ratings than they otherwise would if the outcome assessor had no knowledge of who received what treatment. (This is one of a few ways this kind of bias could occur.) Long story short, outcomes are systematically shifted from the true values, on average, because someone had knowledge of the treatment. Be careful, though as the definition does depend on context (in some cases being as simple as general sampling bias).

thanks so much!
so it is like Observer expectancy bias and Sampling bias?? under what context would Acsertainment bias be Sampling bias?

 

[dempty]

thanks so much!
so it is like Observer expectancy bias and Sampling bias?? under what context would Acsertainment bias be Sampling bias?

When people use sampling bias and ascertainment bias interchangeably it's because they're using "ascertainment bias" to mean how the participants were chosen, i.e. the sampling methodology was non-random. When they're used differently it's that someone is using "ascertainment bias" to describe a bias after the sampling has taken place but during the project (maybe experiment or analysis) that influenced the outcome (i.e. you might have had a random sample, but something else is biasing the results such as the investigator's knowledge of who received what). For exam purposes, I would stick with the context you see in first aid or most commonly in practice questions (which I believe is ascertainment bias as someone knowing the treatments and influencing results unknowingly).

 

[Northerncardinal]

thanks again for your help. One more question: How does increasing sample size affect alpha and power of a study? Based on what i have learnt, an increase in sample size would decrease alpha and should increase power. However, for a fixed sample size, if we decrease alpha, power would also decrease (as probability of beta error increases). How can we decrease alpha for a fixed sample size?