So impressed by this Khan guy, he literally knows everything single shit no matter what you are searching for, his videos will pop up. Thank you very much though :)
I don't understand how the expected proportion can be the proportion calculated from the total of the samples. Could use some more clarification on that. The contingency table is easy to understand though.
Could you explain more precisely the difference of test - Goodness of fit (is distribution the assumed) and test - (In-)dependence of two "things" (contingency table...) of the chi square test? Keep on going.
It seems like this video (and others) focus a bit too much on talking through basic arithmetic. It could be more concisely presented by showing a few examples and fast-forwarding through repetitive content (including repeating what you're writing down because we talk faster than we write). This is especially true in previous videos in this series. Sometimes you say what you're going to multiply to another number 3 times and repeat the result more than 2 times.
Aren't you assuming the herbs don't work when you use 80/380 even though you have no information about it before knowing the result of the test? Seems more logical to me to use the 30/120 ratio to calculate expected values. Could you clarify?
The chi squared test assumes that the underlying distribution is bernoilli right? That's where the divide by expected comes from. Since we are adding bernoillis the variance can be assumed to be the mean, this allows us to use this mean to normalize our errors and get their appropriately normalized normal distributions. I didn't understand the degrees of freedom part. Three normals is three normals right? Why would we use a distribution made of one less normal than the model we have?
Ok, so in this video you were comparing the actual results vs expected results from the placebo and H1 and H2 groups. However, although in this example everything is very balanced and even, how robust is the test against a strong placebo effect or herb effect (in this example). In other words, why don't you compare the data obtained against people getting sick without placebo and without herbs, what if the herbs and the placebo are all protecting people in a statistically significant way compared to naive population?
+noel8421 This is because he is doing the Chi-Square Statistic for Independence...for this type of Chi-Square the Degrees of Freedom are (R-1)(C-1) The previous video he was doing the Chi-Square Statistic for Goodness of Fit...for that type of Chi-Square the Degrees of Freedom are just n-1
Like all the colors, but has no one made a program to find chi on the computer yet? :( And is this used for two-variables only then? I don't really see the purpose of the table, is it to ease our work??
£9000 of tuition fees + hours spent studying = no understanding of chi squared Free video + 20 minutes = complete understanding of chi squared Thank you for the video :)
When our Hypothesis is Herbs do nothing then why include placebo entries? Shouldn't we just compare the two herbs ? If we are to include placebo shouldn't the hypothesis be herbs and place do nothing? What if there was actually placebo effect meaning more people were not affected just by the sugar pill, will the same procedure help in figuring that out?
You're probably done your semester? The placebo allows us to calculate a weighted average. In other words, it tells us how many people in a natural, untreated environment are expected to get sick.
Thank you. It can also show whether there are any effects from either herb that are unexpected and have not been accounted for. Comparing two herbs doesn't tell us anything unless we compare both to something that is expected not to cause changes. In answer to Razikh, the placebo or sugar pills expected effects on a human or animal are already accounted for so there is no unexpected results coming from the placebo. No more than if each persons was given one m and m.
But if our result is closer to 0, doesn't it mean that the real value are actually somewhat close to expected value, which would mean that this was expected and herbs didn't do anything.
It seems to me that three of the pairs are dependent on the other 3. The top row determines the bottom row and vice versa, so one row is redundant, eliminating independence. Shouldn't we use only the 3 number pairs from the sick row? Or maybe the 3 pairs from the not sick row?
Your example doesn't make sense to me... Why would you use the total from all three groups as the comparison distribution? If you're looking for whether the herb data is different from what is expected without them, wouldn't you want to use the placebo group as your comparison distribution? Also, why are you lumping two different herbs into the same analysis? I could understand this if the data were on the same herb from two different flu seasons, or something like that. But it doesn't make any sense to test two different hypotheses (that Herb 1 works and that Herb 2 works) with one statistical test.
goroth01 Yeah, imagine the case where one helps and the other worsen the flu, but when mixed, it cancels out, and we have the ironic result that Herbs do nothing. But we kinda know it is not the case here from the table. He just did not specify the question, maybe in the Question they mentioned Herb 1 and Herb 2 are kinda the same ingredients.
The method that you used to get the expected values, is it different from the method used to get the expected values in a four fold test.. Please could someone answer 😊
if the hypothesis test is herbs do nothing, that means people who are not sick taking the herbs are also sick? So in the end everyone is sick? But in the video the cases are taken the same manner as earlier so what conditions basically the hypothesis put on this?
Shouldn't the hypothesis of "herbs do nothing" be the same as "herbs are placebos"? Is the placebo in this case being the same as some "Herb 3"? Why would "Herbs do nothing" have a relation to the average of the number of sick (and non-sick) people on all 3 categories?
Actually I have no idea, how this video is factually correct, far as I know my text says you get expected values by multiplying say 120 x 80/380 to give 21.05, so what's the logic here? Please remake the freaking video! You don't explain assumptions like how you use 21% for a completely diff column!
You taught me an entire semesters worth of content in a few hours. bless your soul
How are you doing in life nowadays?
@@musicaccount2703 maybe he is dead. maybe you are also dead.
"erbs" :) . Great video
So impressed by this Khan guy, he literally knows everything single shit no matter what you are searching for, his videos will pop up. Thank you very much though :)
10 years later and you still saved my grade :')
Around 4:45 It is said that 80 out of 380 did not get sick but 80 is in the sick row. Or am I not understanding the row label? Confused
I think it is a misspeak
Brilliant! You explain it so well, thank you! I just might pass after all..
I don't understand how the expected proportion can be the proportion calculated from the total of the samples. Could use some more clarification on that. The contingency table is easy to understand though.
Could you explain more precisely the difference of
test - Goodness of fit (is distribution the assumed) and
test - (In-)dependence of two "things" (contingency table...) of
the chi square test?
Keep on going.
In my stats class the professor said NO DECIMALS. Bc you can't have a % of a person. Chi squared is a counting system...
It seems like this video (and others) focus a bit too much on talking through basic arithmetic. It could be more concisely presented by showing a few examples and fast-forwarding through repetitive content (including repeating what you're writing down because we talk faster than we write). This is especially true in previous videos in this series. Sometimes you say what you're going to multiply to another number 3 times and repeat the result more than 2 times.
This video is kind of confusing. I think that more examples would help.
if the herbs do nothing would you not Expect them to have the same %'s as the plecebo, not the totals?
this is still doing mighty wonders. thanks, Proff...
Thanks for this. Just sorted my assessment for this week.
Thank goodness for this channel!
Aren't you assuming the herbs don't work when you use 80/380 even though you have no information about it before knowing the result of the test?
Seems more logical to me to use the 30/120 ratio to calculate expected values. Could you clarify?
Probably the easiest way to calculate expected values is to memorize the formula: (row total * column total) / grand total
Impressive piece of art :)
How did you assume the null hypothesis that herbs do nothing ?
The chi squared test assumes that the underlying distribution is bernoilli right? That's where the divide by expected comes from. Since we are adding bernoillis the variance can be assumed to be the mean, this allows us to use this mean to normalize our errors and get their appropriately normalized normal distributions.
I didn't understand the degrees of freedom part. Three normals is three normals right? Why would we use a distribution made of one less normal than the model we have?
By including the herbs when calculating your expected values, aren't you skewing the data toward non-significance?
I'm learning about this in uni rn, I love to try this in English
Why do you use the total to calculate the expected value? I think using the placebo is more logical (ie. 30/120, 90/120).
I have the exact same question. Good that someone else saw it aswell.
Damn you're life savior ♥️♥️
Where did you got that p table?
I like how at the end it was like "well, we've learned absolutely nothing about the plants" :) great video
Ok, so in this video you were comparing the actual results vs expected results from the placebo and H1 and H2 groups. However, although in this example everything is very balanced and even, how robust is the test against a strong placebo effect or herb effect (in this example). In other words, why don't you compare the data obtained against people getting sick without placebo and without herbs, what if the herbs and the placebo are all protecting people in a statistically significant way compared to naive population?
Hello thank you for this video why is it called contingency table
After all those times you said squaaare during the calculation it sounded funny lol. Thanks for your video!
Thanks, using this as class prep!
Why do you use the placebo measurements in looking at the significance of the herbs? Surely the result from the placebo shouldn't factor into it?
ok... and now he does the d.f. = (rows -1) x (columns-1). Last video he just did n-1
+noel8421 This is because he is doing the Chi-Square Statistic for Independence...for this type of Chi-Square the Degrees of Freedom are (R-1)(C-1)
The previous video he was doing the Chi-Square Statistic for Goodness of Fit...for that type of Chi-Square the Degrees of Freedom are just n-1
Like all the colors, but has no one made a program to find chi on the computer yet? :( And is this used for two-variables only then? I don't really see the purpose of the table, is it to ease our work??
£9000 of tuition fees + hours spent studying = no understanding of chi squared
Free video + 20 minutes = complete understanding of chi squared
Thank you for the video :)
When our Hypothesis is Herbs do nothing then why include placebo entries? Shouldn't we just compare the two herbs ?
If we are to include placebo shouldn't the hypothesis be herbs and place do nothing?
What if there was actually placebo effect meaning more people were not affected just by the sugar pill, will the same procedure help in figuring that out?
You're probably done your semester? The placebo allows us to calculate a weighted average. In other words, it tells us how many people in a natural, untreated environment are expected to get sick.
Thank you. It can also show whether there are any effects from either herb that are unexpected and have not been accounted for. Comparing two herbs doesn't tell us anything unless we compare both to something that is expected not to cause changes. In answer to Razikh, the placebo or sugar pills expected effects on a human or animal are already accounted for so there is no unexpected results coming from the placebo. No more than if each persons was given one m and m.
You need a control in any study like this.
i kinda understand until the end...instaed of using the charts could we just use tehe calculator and find an actual pvalue?
shouldn't you be squaring the numerator AND denominator?
On what condition we will reject the null hypothesis ?
Still pretty impressive. I can't even stay awake for 1 full video...guess that explains my bad grades.
makes sense. the first video was a little clearer however
Sir, thank you for explaining with reasons why we are multiplying the totals🤩I clearly understood
I think i pretty much understand the concept now!
But if our result is closer to 0, doesn't it mean that the real value are actually somewhat close to expected value, which would mean that this was expected and herbs didn't do anything.
I'm just a little confused on this one. Maybe an example in class on Monday?
Pretty good, class review would be magnificent
It seems to me that three of the pairs are dependent on the other 3. The top row determines the bottom row and vice versa, so one row is redundant, eliminating independence. Shouldn't we use only the 3 number pairs from the sick row? Or maybe the 3 pairs from the not sick row?
Great course! More are wanted.
nice work!
Thank you so much
makes sense, but a review would help
-Tyler
more examlpes would help
Your example doesn't make sense to me...
Why would you use the total from all three groups as the comparison distribution? If you're looking for whether the herb data is different from what is expected without them, wouldn't you want to use the placebo group as your comparison distribution?
Also, why are you lumping two different herbs into the same analysis? I could understand this if the data were on the same herb from two different flu seasons, or something like that. But it doesn't make any sense to test two different hypotheses (that Herb 1 works and that Herb 2 works) with one statistical test.
goroth01 Yeah, imagine the case where one helps and the other worsen the flu, but when mixed, it cancels out, and we have the ironic result that Herbs do nothing. But we kinda know it is not the case here from the table. He just did not specify the question, maybe in the Question they mentioned Herb 1 and Herb 2 are kinda the same ingredients.
please just do more examples in class. i like all the colors he used though.
makes sense!
i'm looking for that p-value table and i'm unable to find it. it would have been nice if you'd included a link to it in the description
pretty good
The method that you used to get the expected values, is it different from the method used to get the expected values in a four fold test.. Please could someone answer 😊
Was thinking the same thing, although I'm leaning toward no.....
this makes sense
Thanks, this helped soo much :)
aaarrrrrr, I fell asleep at part 43 of this playlist.
helpful
on some point, you said "21% did not get sick" but you wrote the value under the number of people that did not get sick. so which is it?
Splendid
Makes sense
y isnt the placebo a standard?
Confusing, need more examples for in class.
how did u get 25.3...?
I want that calculator :O
Thanks a very much!! :)
Are you assuming the three groups have the same distribution?
I don't understand how he found the alpha/critical value. Can anyone explain that to me?
People from 2024
👇
sounds goood.
if the hypothesis test is herbs do nothing, that means people who are not sick taking the herbs are also sick? So in the end everyone is sick? But in the video the cases are taken the same manner as earlier so what conditions basically the hypothesis put on this?
a little bit more confusing but i think i get it
why do you say "erb", when its "herb"
sort of confusing. more examples would definitely help. -matt
kinda makes sense
why didn't he just have the total as the expected and the number who got sick as the observed instead of doing all of the other things he did?
thx for the knowledge of 'sugar pill' ... again for briefly explanation of something doesn't related to stats .. .
This example is more confusing
80 out 380 DID fall sick, you did confuse us brah!!
Thumps up so he changes it!
I'm confused earlier in video you said 21% should not get sick if herbs did nothing and then later you say that 21% should get sick
Shouldn't the hypothesis of "herbs do nothing" be the same as "herbs are placebos"? Is the placebo in this case being the same as some "Herb 3"? Why would "Herbs do nothing" have a relation to the average of the number of sick (and non-sick) people on all 3 categories?
this is a little more confusing. more examples would help
really cool videos but you should spend less time using calculator
Um I'm kind of confused why use the 10%?
this would be given in the problem, usually you get 1%,5% or 10% for the alpha, but again it is given in the problem
It prevents you from rejecting the null hypothesis when it's true 90 percent of the time.
im kind of lost listening to him. last video made more sense than this.
@4:44 he says 80 did not get sick but that's the sum of the sick people. Did he say it wrong or did I understand it wrong?
Nvm ignore My comment, he just misspoke he corrects it later.
Do Americans actually say "erbs"?! They sound French!! Great work as usual Sal, sorry to digress.
This is a little more confusing
Actually I have no idea, how this video is factually correct, far as I know my text says you get expected values by multiplying say 120 x 80/380 to give 21.05, so what's the logic here? Please remake the freaking video! You don't explain assumptions like how you use 21% for a completely diff column!
DE ERBZ DO SOMA TING. jus da same as doin nothing tho. ahaha
Still a little confusing
porinchaks videos r better
erbs
grape
sorry for making fun of your voice ha!
this is confusing.
help pervert? xD