as someone who is having to self teach myself statistics, this video (and your overall channel) has been extremely helpful. Thank you for putting so much time into helping others
Hi Justin, thank you so much for all the amazing videos. Just a quick note. Technically I don't think we are allowed to divide infinity by infinity and even if we do, it does not give 1. Probably one needs to think of limits to infinity and use the de l'Hospital rule to say that it becomes 1.
Man you are the best statistic teacher in TH-cam. Please make more new videos. More videos on the different test eg chi square test t test etc. Make a general one then and then how to use each one of them.
Another issue is that frequently researchers are not explicit about what exactly is the population they want to generalize to. Is it all students/classes, all high school math classes, or all public school, traditional lecture style algebra classes in the US? Researchers leave it to the reader to work out what population to "generalize" to by describing their sample (often not very well), but the idea of "significance" for example compares the results from a sample to a specific population. If you just run the tests without setting up the comparisons explicitly, your results could be misleading.
Doesn't it matter how the sample is drawn from the population (random or not)? You're example illustrates this, although you don't mention it. If our sample is starting players, then we cannot be as confident our sample mean will match the population as when the sample is a random mix of players. So, more than just the sample number is important. Also, with regard to the number of the sample (n), sometimes it seems as though confidence goes up with every additional data point, but shouldn't it be gauged by how proportionate the sample is to the whole population? If I have a relatively "large" sample, 1,000,000 individuals, but I choose only Americans and I want to generalize to the global population of 7.8 billion, how good is this sample for making inferences about health, education, or income for example? The standard error might be low just because I have not drawn a random sample and have people who are more similar to each other than the population as a whole. On the other hand, I could randomly drawn 100 people from around the world and that sample would provide better information than the 1 million Americans. If I had 1000 rather than 100, that seems like the confidence would be greatly improved, but it's still a small proportion of the whole 7.8 billion. So I think sample selection and proportion of the population are overlooked criteria for increasing confidence in the generalizability of the inferences.
Is there any methodology on how many significant figures should the descriptive measurements (mean, standard deviation etc) have? Thank you for sharing; this is an excellent presentation.
Nice video, just saw it and helped alot I am doing my thesis proposal and my total population is 400, I wanted an insight on it about which sampling to take please!
If we are in a production environment and we pick 10 out of a 1000 units produced per day, every day. Would the specific question to be answered define what the population is? 10 would be my sample. How confident are we that the sample mean represents the daily mean? Would then mean 1000 is my finite population? Asking for confidence in the sample on a weekly, monthly basis then relate to pouplations of 7.000 and 30.000? Or does the population consist of every unit that ever was and ever will be produced? What if I want to test a certain production process and I do 50 tests. Would 50 then be the population or a sample of an infinite population as one could easily everytime just do one more test? If I want to measure a single piece and remeasure it a lot to get confidence in the measured value would the population then be every measurement that could ever be done?
as someone who is having to self teach myself statistics, this video (and your overall channel) has been extremely helpful. Thank you for putting so much time into helping others
one of the best videos 'by far' narrated in such a humane way ....fantastic
Hi Justin, thank you so much for all the amazing videos.
Just a quick note. Technically I don't think we are allowed to divide infinity by infinity and even if we do, it does not give 1. Probably one needs to think of limits to infinity and use the de l'Hospital rule to say that it becomes 1.
Thanks to Yulin Liu for a note to include a reference to "Sampling without replacement"! Added a few comments to this version. Thanks Yulin!
You are very welcome!
Man you are the best statistic teacher in TH-cam. Please make more new videos. More videos on the different test eg chi square test t test etc. Make a general one then and then how to use each one of them.
Revisiting after many days. Slides do look refreshed.
Content, teaching and method... super, like before.
Sincere appreciations.
That was actually the best statistics video I've ever seen
Man, I just love when this guy says "can't"
By the way, incredible content as always, Justin
Another issue is that frequently researchers are not explicit about what exactly is the population they want to generalize to. Is it all students/classes, all high school math classes, or all public school, traditional lecture style algebra classes in the US? Researchers leave it to the reader to work out what population to "generalize" to by describing their sample (often not very well), but the idea of "significance" for example compares the results from a sample to a specific population. If you just run the tests without setting up the comparisons explicitly, your results could be misleading.
Great video, it was such a great blessing.
Doesn't it matter how the sample is drawn from the population (random or not)? You're example illustrates this, although you don't mention it. If our sample is starting players, then we cannot be as confident our sample mean will match the population as when the sample is a random mix of players. So, more than just the sample number is important. Also, with regard to the number of the sample (n), sometimes it seems as though confidence goes up with every additional data point, but shouldn't it be gauged by how proportionate the sample is to the whole population? If I have a relatively "large" sample, 1,000,000 individuals, but I choose only Americans and I want to generalize to the global population of 7.8 billion, how good is this sample for making inferences about health, education, or income for example? The standard error might be low just because I have not drawn a random sample and have people who are more similar to each other than the population as a whole. On the other hand, I could randomly drawn 100 people from around the world and that sample would provide better information than the 1 million Americans. If I had 1000 rather than 100, that seems like the confidence would be greatly improved, but it's still a small proportion of the whole 7.8 billion. So I think sample selection and proportion of the population are overlooked criteria for increasing confidence in the generalizability of the inferences.
The guy is soooo good. Best place to learn stats
?! Why I only see this video now? Thanks a lot for the great content! Cleared some of my fundamental questions
Outstanding prof and human being
Thanks Justin, really helpful and thankfully you have a voice that doesnt want to lul me to sleep. :)
Thankyou for such an amazing content!
Thank you very much for your amazing video, things get really clear!!!
Extremely helpful. Thank You
Your square root doesn't calculate the same figure in Google Sheets. It says standard dev is 1.052615789. Any idea why?
Well explained. Super helpful! Thank you.
Is there any methodology on how many significant figures should the descriptive measurements (mean, standard deviation etc) have?
Thank you for sharing; this is an excellent presentation.
Same question it will be good to know
Great communicator and teacher
Nice video, just saw it and helped alot
I am doing my thesis proposal and my total population is 400, I wanted an insight on it about which sampling to take please!
how many different samples can be created from a population of size n?
If we are in a production environment and we pick 10 out of a 1000 units produced per day, every day. Would the specific question to be answered define what the population is?
10 would be my sample. How confident are we that the sample mean represents the daily mean? Would then mean 1000 is my finite population?
Asking for confidence in the sample on a weekly, monthly basis then relate to pouplations of 7.000 and 30.000? Or does the population consist of every unit that ever was and ever will be produced?
What if I want to test a certain production process and I do 50 tests. Would 50 then be the population or a sample of an infinite population as one could easily everytime just do one more test?
If I want to measure a single piece and remeasure it a lot to get confidence in the measured value would the population then be every measurement that could ever be done?
Bro hats off to you ...
thanks for this easy to understand
Very Helpful!
Good job dear thnks from Somalia 😍😍
Well done
it was great.thanks
That ∞/∞ = 1 isn't always correct, but it is right in this case anyway.
Never tought i would be interested in statistics
Cool tx sir
But i just wanna to know wht is the reality tells as 0.0471 S E
The estimated mean could be more 0.0471m or less 0.0471m from population mean.
More more more latest videos
I think I've spotted a Sufjan Stevens fan