3 ms·

You need to control to know if the new treatment works, generally. That's my dispute. We know X will kill you in a certain time frame. We have thousands if n

You need to control to know if the new treatment works, generally. That's my dispute. We know X will kill you in a certain time frame. We have thousands if not millions of data points. I fail to see how the control group is really telling you anything you do not already know. Ignoring what we already know is unethical. It is just is insane that we set people up for false hope and watch them die in droves. I guess it looks better on the marketing material to doctors. I wonder how many actual studies are halted because of die off in the control group. It sure seemed like that wasn't going to be an option.

jcranmer 7y ago
One of the reasons for a control group is to provide opportunities for blindness and double-blindness. Placebo effects are (hopefully) very well-known, but that's not the only common cause of systematic error. Allegiance effect also exists--a doctor may score the same patient with the same symptoms differently if there is a reason to favor one interpretation over another, and it need not even be a fully conscious effect.
jrochkind1 7y ago
> That's my dispute. We know X will kill you in a certain time frame. We have thousands if not millions of data points. I fail to see how the control group is really telling you anything you do not already know. I don't think you understand the purpose of the control group. It is irrelevant how much we already know about what Alzheimer's disease does, because the control group is not intended to tell you anything about that at all. The justification for the control group is not that it will tell us something about what Alzheimer's does. The purpose of the control group is to determine if the treatment works. That's it. You seem to be suggesting that when doing a research study of a new treatment, it would be unethical not to give this treatment to everyone on the planet who has the disease. That makes no sense at all. The experimental treatment may do nothing for them, or may actually harm them. Giving an experimental unproven treatment to everyone would be unethical. If you want to argue that it is unnecessary to have control groups in an experiment, you need to first understand what the purpose they are said to serve by those who believe they are important (which is pretty much any statisticians or methodologists).

jcranmer 7y ago

One of the reasons for a control group is to provide opportunities for blindness and double-blindness. Placebo effects are (hopefully) very well-known, but that's not the only common cause of systematic error. Allegiance effect also exists--a doctor may score the same patient with the same symptoms differently if there is a reason to favor one interpretation over another, and it need not even be a fully conscious effect.

jrochkind1 7y ago

> That's my dispute. We know X will kill you in a certain time frame. We have thousands if not millions of data points. I fail to see how the control group is really telling you anything you do not already know. I don't think you understand the purpose of the control group. It is irrelevant how much we already know about what Alzheimer's disease does, because the control group is not intended to tell you anything about that at all. The justification for the control group is not that it will tell us something about what Alzheimer's does. The purpose of the control group is to determine if the treatment works. That's it. You seem to be suggesting that when doing a research study of a new treatment, it would be unethical not to give this treatment to everyone on the planet who has the disease. That makes no sense at all. The experimental treatment may do nothing for them, or may actually harm them. Giving an experimental unproven treatment to everyone would be unethical. If you want to argue that it is unnecessary to have control groups in an experiment, you need to first understand what the purpose they are said to serve by those who believe they are important (which is pretty much any statisticians or methodologists).