Confidence Interval: The right and wrong way to understand them.

[@datatab]

If you like, please find our e-Book here: datatab.net/statistics-book 😀

[@IrishMexican]

As a data analyst on a data science team, you make me feel less dumb in my position. Thank you.

[@kennethgottfredsen767]

The best source of learning statistics on KZbin.

[@datatab]

Many thanks again : )

[@putin_navsegda6487]

thank you, you explained it really well !

[@datatab]

Glad it was helpful! : )

[@ramyan5528]

Nice explanation!!! . In simple words we can say, Confidence Interval tells about the precision not the accuracy!

[@datatab]

Yes! Many thanks for the addition! Regards Hannah

[@cantkeepitin]

This sentence is even more confusing than CI's.... Indeed parameter estimation method may also suffer from bias errors.

[@SnehalPatel-dc8gk]

At 2:12, you start off correctly in saying that the explanation is not correct "when viewed from a frequentist statistics perspective"; however, that doesn't make it wrong.

[@datatab]

Hello, thank you for your feedback. What exactly do you mean that? Sorry, I don't know if I understood the point correctly and would be happy to receive a clarification! Regards Hannah

[@chanrasmeypech9710]

I like your channel. I do learn a lot from this. Thanks for your kind sharing the knowledge. Very much appreciate ❤

[@datatab]

Many thanks for your nice feedback!!! Regards Hannah

[@heliomonteiro4200]

Great vídeo, thanks

[@datatab]

Many thanks! Regards Hannah

[@caduguimaraes]

Wonderful explanation

[@datatab]

Glad it was helpful! Thanks and Regards Hannah

[@gordoaraujo]

Great video! What is the difference with confidence bands? Do they also have similar interpretation? Thanks :)

[@datatab]

Many thanks for your feedback! Confidence bands apply to a function or curve (e.g., a regression line). Instead of a single interval at one point, confidence bands give you a range of plausible values for the entire curve. Regards Hannah

[@mustafizurrahman5699]

Excellent

[@datatab]

Thank you so much 😀

[@MarkFrankUK]

The difference between A and B raises interesting issues in the philosophy of probability but I am not sure it makes much difference in practice. You would lose money in the long run if you regularly bet on the true value being outside the 95% confidence interval at odds of 10 to 1.

[@Septounze]

Good stuff!

[@datatab]

Glad you enjoyed it and many thanks for your feedback! Regards Hannah

[@lpdowdell]

do you sell your book in Amazon?

[@datatab]

No sorry, jsut as pdf on our webpage!

[@Questtoknowwithdraftab]

Nicely explained

[@datatab]

Many thanks! Regards Hannah

[@mohammedelbarbary8708]

Could CI contain negative values?!

[@datatab]

Yes this could be!

[@PassingTheDog]

Thanks :)!

[@datatab]

Welcome! : ) Regards Hannah

[@levu74]

What does the parameter mean?

[@datatab]

The parameter is just for example the mean.

[@digishah]

So if I did an experiment to calculate the mean height for a sample of boys vs girls with 95% confidence interval. Using the confidence interval to determine whether their height difference is significant would be wrong?

[@MusicIsLifeWithGuitar]

Hi Hannah, thank you very much for this video. It was really helpful and I realized I was interpreting the CI as explanation "A". But using the true definition as "B", how would we then interpret the confidence interval on a normal basis when we usually only draw one sample to represent the true population. Can we even make an interpretation of the confidence interval or even use it then when we only draw one sample?

[@johnkalas]

Yes, we can. The confidence interval (CI) is based on the single sample. However, the interpretation of the meaning of the CI does not involve a probability about the single sample, but rather a probability with respect to the vast number of CI’s which are possible, given the strategy that was used to produce this particular CI. The many different samples which are possible each produce their own, various CI’s. We are saying that, if we repeatedly use the same method that we used to produce this confidence interval, then 95% of the time, we would expect the CI that gets produced to contain the true value of the parameter. In fact, we can make a statement about the meaning of the confidence interval BEFORE we take the sample: “I am going to use a certain method to construct a CI for the population parameter. If I were to use this method many times (which I am not going to do), then the relative frequency of the event that the resultant interval contains the true value of the parameter would be 95%.” Again, the probability is a reference to the method being employed to construct the CI, not the single CI which results from its use.

[@kennethgottfredsen767]

Are you going to cover any machine learning and Bayesian statistics in the future? Or are you sticking to the basics of statistics?

[@datatab]

Many thanks for your comment! hmm, we haven't fully decided yet, maybe next year we'll try out going a little further! Regards Hannah

[@yadukrishnan8577]

🤦‍♂9:25 "How is the confidence interval for the mean calculated?" -- This statement is confusing to me. Isn't it better to say, 'How is the confidence interval for this method/statistical-test calculated?'

[@djardar]

Ok, say I have performed some experiments, and I have a confidence interval. It cannot give me the probability that it contains the true mean. How is that useful at all? Why bother?

[@datatab]

Hi, when you say you have a 95% confidence interval, it means that if you repeated this experiment many times, 95% of the intervals you calculated from those repeated experiments would contain the true mean. It doesn't tell you the probability for this specific interval, but it does tell you how confident you can be in the process of estimating the mean. It's not about the probability for one interval, but about how much trust you can have in the method you used to generate that interval. I hope this was helpful, Regards, Hannah

[@djardar]

@@datatab Thank you for the swift reply. I still struggle with how they are useful; is the main goal to compare them to each other, since it doesn't like one interval by itself says much (besides how much you trust the sampling method)? What if you look at historical data, and group by months for example? Then you could have one CI per month..

[@charlesmanning3454]

In the long-run aren't we all frequentists? The difference between saying there's is a 95% chance the true value of the parameter lies in the CI and saying the method will produce a CI containing the parameter 95% of the time, seems small and of little consequence, if there is one at all. You don't know the true value of the parameter so you take a sample. You know the method you used will produce a CI containing the true value 95% of the time. If you would have taken a 100 samples then about 95% of them would contain the true value. If you took 70 samples about 95% of them would contain the true value. if you took 10 then 95% of them .... But you only took one so why is it wrong to say there is a 95% chance it contains the true value?

[@DebEternity]

It's not the correct interpretation that out of all the samples, 95% of them contain the true value (even though it is actually true). Rather, the correct interpretation would be if a sample is selected at random, there is a 95% chance that it will contain the true value. The key difference in interpretation here is that the probability rests on you making a choice at random, not in the existence of the samples by themselves. For instance, if there's a bag containing 100 balls, with 95 red balls & 5 yellow balls, it would be accurate to say that there's a 95% chance of you picking up a ball at random & it being red. It would also be accurate to say that 95% of the balls in the bag are red. However, the first one is a probabilistic event that is yet to happen & therefore a prediction of an outcome is made. The latter is not a probabilistic event but rather a statement about the contents of the bag. A 95% chance of you picking a red ball from the bag is not the same sentence as 95% of the balls in the bag being red. I think the manner in which we interpret percentages & the way we relate it to probability is throwing us off. I hope I was able to explain my thoughts clearly. Cheers!

[@ulyssesfewl1059]

Maybe I am just dumb or stupid or both, but this makes no sense to me, sorry.

[@myworldfriends123]

😀😃🤩🤩🤩🤩🙂😊

[@datatab]

Thanks : )