I'm in college now and have been watching videos from this channel since high school. Needless to say, they've been so helpful. So crazy and great I can find a resource as great as this on RU-vid!
taking 4000 level biostatistics and this is much more effective, and efficient, explanation of hypothesis testing! I feel like I wasted hours reading a book that is written in archaic jargon, when everything I needed to understand the logical concepts is right here.
*_...sometimes statistics talks so slightly about usability, that it looks offtrack from any-use-at-all (mathematics isn't to blame)-When I consider votation outside the entropic zone of indistinguishability from random-voting, I include the likelihood of rejected results being returned for a specific reprocessing (e.g. contested Laws returned to the Supreme Court whose workload is fixed), and I perhaps-arbitrarily for otherwise-general use select σ√2π as where the 'Normal' bottoms-out..._*
Could you please talk about effect size. And if it’s as important as statistical significance. Especially with complete population stats (like census). So we divided the us population into males and females, can we use effect size?
Ytterbium-171 atomic clocks demarcate a second with their transition frequency of 518, 295, 836, 591, 614 (518 trillion and change), with an Allan's deviation of 10^17! And you thought that you had control over your p.
Echoing the comments about food...Food serves as a such a perfect source of relevant and manipulatable examples for a stats course. Three thoughts: 1) She does a great job using food in just this way in every one of the videos I've watched so far (up through 21) 2) People with eating disorders should look to Khan Academy for their video-based stats ejumukashun. Though inferior to CC in myriad ways and overall, at least his chicken scratch drawings and overly assumptive examples in no way resemble something comestible. His videos rival unadorned chicken breast in blandness and enjoyability. CC videos, well, chicken parm baked by a Neapolitano grandma named Maria! 3) She's an obvious plant by the Koch brothers and big-ag.
I used program to generate a normal curve stated in last problem with mean of 2300 and SD of 500/sqrt(60) to get p value of 12% and not 18% as said in the video. CC please expaln?
No. You should expect statistically significant findings to be incorrect more than 5% of the time, because statistically insignificant findings are very rarely reported. Also keep in mind that you can sample your population 20 times, for which you can expect 1 significant finding. To counteract this, you correct for multiple testing by controlling the family wise error rate or false discovery rate.
You take a look at the distribution of your null hypothesis and calculate the area under the line from the value of your test statistic towards infinitely more extreme. The area under the the line of the entirety of your null hypothesis distribution should add up to exactly 1, thus the area that you are calculating for your statistic has to be smaller than 1.
I'd really like to love this series, like the many excellent other subjects, but it doesn't quite have the same flair. Disappointing to see logical contradiction compared to unlikely.
Can someone explain to me this conclusion 8:08 ? : I understand that 9% of mean caloric intake are higher than 2400. And that since the distribution is symmetric, 9% higher than 2300 is 9% lower than 2300. But how does they add up ? WHY THE HECK do we care to use the symmetric rule to say that 18% are away from 2300. Where does 2400 fit in this conclusion ??? I need help seriously y_y
The p-value is the probability of the data randomly occurring. If the p-value is high, then the data is considered not significant and you fail to reject the null hypothesis. If the p-value is low, then the data is likely not random and you have statistically significant data (you fail to reject). The p-value for this problem is .09, while the standardized p-value cut off in statistics is .05. In this example, the probability of the data randomly occurring is 9%, so we can't say the data is significant since the probability greater than 5%.
@@freezerburnfreeze Thank for you answer although that is not what I wanted to understand, I actually do understand well the p-value but I don't get how a two sided p-value work in this example. They use the symmetric rule but it is not the logic behind it that I don't understand, it is how do they manage to use it here that is not clear.
In this case, the question is whether people with gene X eat a different amount of calories than the general population (NOT whether people with gene X eat HIGHER amount of calories). Although we got the information that the mean of a sample of 60 people with gene X is 2400 and is greater than the mean of general population which is 2300, the mean of the all the people with gene X might be smaller or greater than 2300, so we need two cutoff numbers on both sides to decide whether to reject NHST and it's .025 for each side, and comparing .09 to .025 is just same as comparing .18 to .05.
I hate to nit-pick, but your examples misused standard deviation. 175 spots with a standard deviation would have a bell curve MUCH wider than what is shown. The bell curve shown has a mean of 175 with a standard deviation of approx 11 or 12. Please make your graphics match your examples.
the graphs in this example use the standard error, not the standard deviation since it's the mean number of spots for a group of giraffes not just one!
You say that the current consensus is that alpha should be 0.05 - this is standard in the social sciences. But ask a physicist? When they found the Higgs boson they went out 5 standard deviations from the null hypothesis mean.
She could not have been more unclear in her definitions and examples. She literally made it as complex and confusing as possible even though this concept is actually not so.
There are so many things wrong in this video I can’t even start to describe. Just get a statistician to take the show and correct the misconceptions, please.
Stating that chance alone cause a test to be wrongly significant and use of hard cutoffs are the most common and dangerous misconceptions here. Just read the ASA statement about p-values (February 2016 if I remember correctly) which talk about these issues.
@@BernardoPowaga There is a whole episode in the serie about that p-value "standard" discussion and the reproductibility crisis in sciencitfic research, tho. And she does mention the ASA statement and provide a link to it.