Talk:Censoring (clinical trials)

Does reducing sample size always reduce reliability?

"Censoring a patient will reduce the sample size for analyzing after the time of the censorship. Reducing the sample size always reduces reliability . . ."

This assertion is unsourced, and it strikes me as dubious. Sample size is a critical factor in statistical reliability, but there are other factors as well. If there are confounding issues with a member of the sample, or something has happened to bias the results for a particular patient, it may make sense to remove them. Perhaps a study was intended to be double-blind, but something has happened to inadvertently reveal to a patient whether they are a member of the study group or the control group. I'm not familiar enough with the process to know all the reasons a researcher might choose to censor a trial, but I find it difficult to believe that it's always making the study worse.

Perhaps the intent was to say that it always reduces the statistical reliability of the study; if so, that should be clarified, and probably better words should be used. You can calculate a confidence interval based on a given sample size, but that assumes a lot of things, such as independence of the sample observations. If you're censoring because one of those assumptions is violated, you're not actually reducing the reliability of your results. EastTN (talk) 16:12, 27 October 2008 (UTC)