Home › Forums › Decaffeinated Coffee › Things Causing Autism (can only be a) Joke › Reply To: Things Causing Autism (can only be a) Joke
Stam
Continuation of #2 above.
In order to understand why the study you demand is impossible we need to discuss (briefly) an overview of study design.
you begin with a theory that you want to test, say Vaccines cause autism (or more accuratly are linked to autism) Or lets take a simple example you have a coin you suspect is an uneven coin.
You know want to design an experiment to determine if your coin is uneven (lets assume an even coin has a 50/50 shot of landing on heads) so you set up a “null hypothesis” That is that there is no difference between your suspicious coin and a regular coin. Our experiment will be designed to determine whether we should accept or reject the null hypothesis in favor of the “alternate hypothesis” which in this case would be that the coin is uneven.
Now obviously simply flipping the coin once and getting heads doesnt tell you to much. nor does flipping it twice even if both are heads, after all flipping an even coin twice would be expected to result in 2 heads 25% of the time This hardly proves the coin is uneven. Obviously you will have to flip your coin several times.
Now lets take it to the other extreme. Say you flip your coin 100 times and all 100 are heads. Is it impossible to have occurred by chance alone? Of course not, while admittedly suspicious it is POSSIBLE that you have an uneven coin and the unlikely occured. What to do with the hypothesis? To make it more complicated what if the split was 90 heads to 10 tails or 70 to 30 or 55 heads to 45 tails. Which of these splits are “statistically significant” proving that your coin is uneven and which are expected occurrence of chance?
To answer this question a “p value” is needed. The p value is a calculation of how likely the outcome is to occur based on chance alone. Generally a p value less than 0.05 or 5% is used. This means that in any significant study with a p value less than 5%, there is a less than 5% chance that the outcome occurred based on chance alone. By flipping a coin 100 times yielding 90 heads the p value is 0.0045 (I totally made this up, maybe charliehall can help if hes bored) this means that there is a 4.5% chance that the outcome occurred randomly which leads us to reject the null hypothesis and conclude that the coin is uneven.
So far so good?
Ok now on to your study