Another biostats tidbit - confidence interval

[Phloston]

---

Members don't see this ad.

If we increase n, then we increase power, and the CI becomes narrower. This means the study becomes more precise.

What about accuracy?

 

[withrye]

Accuracy is more dependent on the study design. If you're trying to assess the general prevalence of breast cancer, but your study consists of 200,000,000 males, you're gonna have a very precise, very wrong (inaccurate) answer.

 

[mmmcdowe]

The confidence interval is more related to alpha (indirectly) and the size of your standard deviations than power, but yes confidence intervals do decrease with increasing n. However, this isn't the definition of precision. Precision is more related to your standard deviations of your sample. If your sample is all over the place (even if in reality the population standard dev is small) because of poor quality measurements then the high standard deviation will increase the size of your CI.

Assuming your methods are perfect, increasing sample size will increase accuracy because as n increases it reaches N, the total population size. Once n=N, your accuracy is 100%. However since methods are rarely perfect, you won't ever reach a true 100% accuracy, but it still will increase with n.

 

[Phloston]

Thanks a lot. Once again, very helpful.

 

[mmmcdowe]

Thanks a lot. Once again, very helpful.

My pleasure.