AMS 5 Lecture 2: Class 2 - Controlled Experiments

• School: UCSC
• Department: Applied Math and Statistics
• Course: AMS 5
• Professor: Yonatan Katznelson

Document Summary

Randomized controls tend to achieve this goal efficiently (and more accurately): if the treatment and control groups differ significantly in aspects other than the treatment itself, then it is more difficult to assess the effect of the treatment. Example: randomized controls vs nonrandomized controls in portacaval shunt studies. In the poorly-controlled portacaval shunt studies, the confounding variable was the health of the subjects. Who is eligible: step 2: assign them to the control/treatment groups. Four therapies were evaluated both by randomized controlled trials and by trials using historical controls. Conclusions of trials were summarized as positive (+) about the value of the therapy, or negative (-). 0: different experiments conducted to test the same therapy can lead to opposite conclusions depending on the design. In a controlled experiment, the researchers decide which subjects are in the treatment group and which subjects are in the control group.