Double blinds are good, but triple blinds are better. The patient may have the placebo effect, the researcher may unwittingly indicate the type of treatment on way or the other, and the analyst may select data based on what they expect to see. The placebo effect is a well-documented phenomenon. The researcher may have bias, but one issue that I haven’t seen raised is that they may signal the patient that they are being treated on way or the other. This would then start the placebo effect. This all would skew the data towards one outcome over the actual results. When the analyst doesn’t know what he/she is looking at, they can sift through and analyze the data without imparting their own bias onto the results. When removing outliers or scaling results, they cannot change the data to fit a certain outcome if they don’t have an idea of what the outcome should be. The more blinds, the more assurance the study has of being impartial and removed of human bias towards a certain outcome.