Timeline for Flaw in all scientific studies/experiments?
Current License: CC BY-SA 3.0
18 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Oct 24, 2016 at 15:51 | history | edited | user127039 | CC BY-SA 3.0 |
This is why random assignment doesn't make distribution of participants equal
|
| Oct 17, 2016 at 20:07 | vote | accept | CommunityBot | ||
| Oct 6, 2016 at 2:35 | history | edited | user127039 | CC BY-SA 3.0 |
edited title
|
| Oct 5, 2016 at 21:19 | comment | added | Aksakal | This is not science. Look at how they measure "energy". It's some kind of a qualitative approach in "cargo cult science" type of "research". Once the folks start borrowing precisely defined terms like "energy" from physics, it's a good sign that the folks are lost. It's sad actually. | |
| Oct 5, 2016 at 19:24 | comment | added | user127039 | @DJohnson what did you mean by, "are true in the limit"? I'm not sure what you meant by 'limit' in that context. On first read I believe you're saying that the assumptions in random trials only apply to an infinite data set, and don't hold in real life. | |
| Oct 5, 2016 at 18:22 | comment | added | user78229 | Regrettably, many of the assumptions involved in random trials are true in the limit but don't hold for finite samples of data. | |
| Oct 5, 2016 at 16:48 | answer | added | Matthew Gunn | timeline score: 19 | |
| Oct 5, 2016 at 16:37 | comment | added | user127039 | @dsaxton okay, so you can pretty much draw the conclusion that science proves nothing.. which isn't a new concept, it's just heavily abused in politics.. | |
| Oct 5, 2016 at 16:13 | comment | added | dsaxton | Science deals in probabilities and not certainties anyways. No one believes treatment groups are 100% balanced in every way, or that one experiment proves anything. | |
| Oct 5, 2016 at 16:05 | comment | added | user127039 | @dsaxton It suggests that causal conclusions are flawed because it relies on the assumption that the imbalance is minimal. Random specifically does not guarantee minimization. Random means you can get 50 heads in a row and only 1 tail from a total of 51 flips. All those heads were as equally likely to happen as tails. But in the end you ended up with a bunch of heads.. You ended up with worse case. The only way to be slightly more confident you didn't make false conclusions do to screwed treatment and control groups is to repeat the experiment dozens of times.. | |
| Oct 5, 2016 at 16:02 | comment | added | JasonD | It's perhaps a bit alarmist to suggest all experiments are flawed. In cases such as these, where people are involved, authors will sometimes report the characteristics of their groups post hoc to demonstrate how balanced they were (or not). I agree that randomization alone doesn't guarantee balance, but as @dsaxton points out, it's what we've got and is helped by having larger n in each group. | |
| Oct 5, 2016 at 15:57 | history | edited | user127039 | CC BY-SA 3.0 |
adding space
|
| Oct 5, 2016 at 15:54 | comment | added | dsaxton | How does this suggest science experiments are flawed? Of course you will end up with slightly unbalanced treatment groups, but the point of randomization is that the imbalance is minimized to the best of our ability. | |
| Oct 5, 2016 at 15:50 | comment | added | user127039 | @Björn but to answer your question more directly, the example compares the two, and I'm asking about the flaw in the randomized study-- it doesn't make sense to conclude any thing causal due merely to the logic of, "well, we randomly assigned them to groups, therefore we can make a causal conclusion because people who have high energy to start are likely equally represented in both groups".. there's no guarantee that they are.. | |
| Oct 5, 2016 at 15:46 | comment | added | user127039 | So, the point of this particular video was to basically say, "hey with observational studies, you can only make associations, and not causations, but with an actual science experiment, where you randomly assign people to one group to work out, and one group to not work out, you can make a causal conclusion, because such variables that might also contribute to the outcome (like already having high energy level to start) are likely equally represented in the two groups due to the random assignment" and my point is , "likely equally represented" doesn't guarantee equal representation: it's random | |
| Oct 5, 2016 at 15:38 | comment | added | Björn | What is the example about? A randomized or an observational trial??? | |
| Oct 5, 2016 at 15:34 | history | edited | user127039 | CC BY-SA 3.0 |
added 214 characters in body
|
| Oct 5, 2016 at 15:29 | history | asked | user127039 | CC BY-SA 3.0 |