When alchemize the ingredients, we shave off balanced portions, and are then left with the unbalanced remains.
If we didn't alchemize at all, the ingredients would approach a more balanced relationship over time.
When alchemize the ingredients, we shave off balanced portions, and are then left with the unbalanced remains.
If we didn't alchemize at all, the ingredients would approach a more balanced relationship over time.
Dude, you are clueless.
The expectations are know ahead of time; there is no more or less variability (as the data gets larger) as a percentage of the data. You should have learned this concept in basis statistics.
Here is a newsflash for you: Everyone’s results fall within expectations.
Your first claim of “Remember probability indicates over time, any outliers should be minimized“ is a very STRONG claim and is false because the (expectations of the) percentage of the outliers are INDEPENDENT of the sample size, e.g. the number of data (when assume the distribution to be normal).
Your second claim of “says the outliers relevance will be getting minimized as they become a small part of the total outcomes” is equally wrong for the SAME REASON.
Here are two Wiki sites to help you understand the area under the curve (for a normal distribution), where outliers are 4-sigma, 5-sigma, etc events.
https://en.m.wikipedia.org/wiki/Standard_deviation (please see the last table in the Intepretation & Application where it shows the percentages for 1-sigma, 2-sigma, 3-sigma ... up to 7-sigma)
https://en.m.wikipedia.org/wiki/68–95–99.7_rule
The area under the curve for the mean +/- 1-sigma or 1 Standard deviation is roughly 62.87%
This is true if you had 10 data points, 30 data points, 100 data point, etc. This roughly 62.87% (a percentage) is INDEPENDENT of the number of data points.
The area under the curve for the mean +/- 2-sigma or 2 Standard deviation is roughly 95.45%
This is true if you had 10 data points, 30 data points, 100 data point, etc. This roughly 95.45% (a percentage) is INDEPENDENT of the number of data points.
The area under the curve for the mean +/- 3-sigma or 3 Standard deviation is roughly 99.73%
This is true if you had 10 data points, 30 data points, 100 data point, etc. This roughly 99.73% (a percentage) is INDEPENDENT of the number of data points.
This also applies to 4+ sigma events, e.g. the outliers. These outlier PERCENTAGES are INDEPENDENT of the number of data points.
Had you understood logic, if the percentages are INDEPENDENT of the number of data points, then how does the outliers get MINIMIZED? It makes no sense. Your assertion is there is a negative correlation between outliers and data points; as you get more data points, you get fewer outliers which is contrary to basic statistics.
In summary, since the percentages of outliers is INDEPENDENT of the data points, the revelance of the outliers (as a percentage) DO NOT CHANGE, and if there is NO CHANGE, how do they get minimized (as percentage)?
In Statistics, the term we use is Survivorship Bias.
If we only look at the current or remaining ingredients, it looks unbalanced due to survivorship bias. Therefore, the correct way to look at the drops is the total drops to date and compare that number to expectations. Unfortunately, we know from statistics some people will be facing a ton of negative variance.
My favorite example of this is the "proof of miracles" story.
We hear plenty of stories about how a survivor faced almost certain doom, prayed, and his prayer was answered.
What we don't hear are stories from the other 99% that prayed but their prayers weren't answered -- because dead men tell no tales.
Just an FYI ... two ways to quickly debunk cat13’s assertions but these explanations requires a good command of logic:
#1 If the large of law numbers tells us we approach expectations as the data points / observations get larger, then that must also apply to outliers. As we get more data points, the outliers approach their own respective expectations (vs cat13’s claim of it gets minimized).
#2. This argument requires using the concept the whole is the sum of its parts. If the whole is made of non-outlier data (say 3-sigma or less) and outlier data (anything over 3-sigma like 4-sigma, etc), then for cat13’s assertion to be true that outliers get minimized, then the non-outlier data MUST get larger as we get more data points, respectively.
Unfortunately, cat13 doesn’t understand the percentages associated with nonoutlier data is immutable. Logic dictates if one part of the whole (the percentages) is immutable, the remaining part of the whole (the percentages) is also immutable.
You might not like his wording, but I don't see that Cat13 is making any claim the numerical results change.
He's just saying that as they accumulate, the imbalances aren't perceived as extreme relative to the whole data.
I would agree, but we don't allow the data to accumulate -- we shave off balanced portions to alchemize every time we get a chance -- and for that reason the extremes are left in our inventory.
Or we can just give Kobo those 30 troll tusks...