If you had a positive result from a study of two out of a population of 7 billion, what are the chances that it would be a fale positive/statistically insignificant?
Sample size = 2?
>>8222208
Yes, sorry.
What I'm asking is:
If a study with a sample size of 2 has a positive result and the population the sample is drawn from is 7 billion, is there anyway to calculate the probability of the positive result being insignificant?
Or calculate how likely/unlikely it is to be accurate?
>>8222204
A very overwhelming number. A sample of 2 out of 7*10^9 is [math]nigh[/math] insigificant in terms of sheer quantity.
>>8222235
What does nigh mean?
And I know it's insignificant, but is there a way to work out how insignificant?
>>8222204
Assuming that your samples are representative of the total population, the probability that it is a false positive is equal to the p-value. If it is a positive result, then by definition it is statistically significant. Calculating the false discovery rate only becomes possible if you do multiple statistical tests, for instance if you run the experiment again and with a different sample.
Counter to what many people think, the N is not the critical factor per se, but what matters is the amount of data per case in the total N.
>>8222235
No, learn to stats.
>>8222250
>Calculating the false discovery rate only becomes possible if you do multiple statistical tests, for instance if you run the experiment again and with a different sample.
Good point.
But how do people determine the necessary sample size for an effect that is intended to be measured?
>>8222243
If you can not get bigger sample you can use any strict similarities but w/o any guaranty. You can guess if some variables continuous using other science but it is not about exploratory data analysis
>>8222255
You can calculate the power that you need, and it depends on the expected size of the effect. The weaker, the more data you need to factor out the noise. The effect size is usually estimated from the existing literature.
If you actually want to do this, just google 'power calculation'.
>>8222243
It depends on the research. Let's say your study is about verifying whether people need to have their eyes open as part of enhanced physical development and in consequence of this, better chances of survival in environment.
Take 2 people out of 7*10^9 to do this research. Make them do things such as close their eyes, look at the light, look at mirrors, read a line of text and so on. You reach the conclusion that open eyes serve of enormous help to these two humans and for you as well. The chances of it being helpful to all humans should be proven with correct data, since all humans beings are basically constituted of the same sensorial abilities to sense the world.