Hypothesis testing. Null vs alternative

[@365DataScience]

🚀Sign up for Our Complete Data Science Training with 57% OFF: bit.ly/3sGBk7a

[@aaron-fo2gc]

Probably the clearest explanation ever, that "Innocent until proven guilty"

[@annacubito46]

YES SAME OMG

[@hanaeldeeb9280]

YESS

[@austinkim9402]

Therapist: "Skinny Trump doesn't exist. He can't hurt you." Skinny Trump: 1:08

[@dylanstevens4589]

The greatest President of all time

[@just2581]

I was taught to never accept a hypothesis, but to either reject it or not reject it

[@applepeel1662]

Daaamnnn same

[@harineepriya]

Same

[@layaunabonds2869]

this all my professor had to saaaayyyyy! THANK YALL

[@Prosperous903]

lucky to have such lecturer

[@ChiTwnCutie]

GOT IT!! THANK YOU!! Clear and concise! Excellent examples!!

[@365DataScience]

Glad it helped!

[@Cheyennep]

Thank you for helping me with my psych statistics class😭

[@cockerspaniel1973]

Your awesome, knowing something and breaking it down so others can comprehend the concepts and mechanics are vastly different. You are an instructor and a good one.

[@sumayyamazhar3422]

Ohhh i really want to get it inside my head. But due to continuous studying i'm a little dizzy and feel like i've had enough booktime today.

[@charlulu167]

Rest

[@Pulmonary.Tuberculosis]

Thank you so much. I didn't quite understand the explanations in my school textbook, this video really helped.

[@deemaaladhadh6306]

THIS MADE ME MORE CONFUSED

[@aravindu6403]

Finally someone has said it

[@jexdiosan5851]

Yes confused all the more because I was thought that null hypothesis is negative or indifference, while alternative hypothesis is positive.

[@mariakathleenkerstinealiya3130]

@@jexdiosan5851 what's the difference between them tho?

[@pandanope7846]

@@jexdiosan5851 That is what I have read too. "Null" meaning "non existent" or "amounting to nothing".

[@harshakumar8970]

Thank you for turning on/adding Closed Captions!

[@365DataScience]

No problem!

[@KhalidMahida]

Excellent video, clear concept in just 4 minutes

[@anajudith7377]

THAAAAAAAAAAAAAANKS! I was really struggling understanding this concept.

[@jamesholmes4381]

The example @3:20 is incorrect. If I recall correctly, the alternative hypothesis cannot contain an equals sign. the equals sign is always given to the null hypothesis

[@AustinCS]

This would be an accurate statement. The video is wrong in this example.

[@mikasan237]

I love the example given for what a testable hypothesis is :)

[@alessandroconcetti3605]

I'm confused... If I had to test my hypothesis "data scientists' salary is greater than 125,000", I would say that my null hypothesis is "data scientists' salary is less or equal than 125,000" and with a statistical test I would check if I have enough confidence to reject the null hypothesis and prove my original hypothesis. In my head, rejecting the null hypothesis is a good thing. Isn't it?

[@susanfinch3209]

I was looking for someone to say that. The null should always contain equality.

[@mreagantn]

Very good explanation. Thank you!

[@365DataScience]

You are welcome!

[@ritzie7611]

Wow, what s clear and concise explanation.thank you!

[@akoredeadebayo4269]

bless your ♥, it finally makes sense

[@williamofy6376]

By definition, the NULL H0 hypothesis is the ONLY hypothesis that can have an equal sign...you are wrong in your explanation here @ 3:27...So it should H null: Mu is less than or equal to 250,000...

[@TenorSaxx]

In example 2.0 there is a critical mistake. The null hypothesis MUST ALWAYS contain some the equality. Therefore the null and the alternative in example 2.0 should be flipped. This would make you lose some points on your tests and might confuse everyone. Nice video otherwise but this makes me question everything else I've seen in the video.

[@syamilysasi8677]

Sir could you please explain? I thought what was shown in the presentation was correct. In the example, he is trying to test or disprove his friend's argument. So he made null hypothesis(which is always made to reject or nullify) based on his argument and made alternative hypothesis based on the negative of that argument. Is it not correct?

[@TenorSaxx]

​@@syamilysasi8677 So the null hypothesis is different for different tests (P-Value, T-test, Chi-Square test, ANOVA, etc) however the common underlying premise is that the null hypothesis presumes that there is no difference between the new data observed and the previous, old collected data; that there is "equality" between the two. It is called "null" because it is assumed true; it is upon the test to prove that it isn't (Not the best analogy but imagine the null as a person in a law suit: you are innocent until proven guilty. It is upon the evidence of the observed data to prove your guilt). To test, we have to show that the known data and the new observed data have a "significant" difference. This difference is somewhat arbitrarily assigned as a measurement of variance, commonly two standard deviations from the mean (highest point of the bell curve). If our observed data is outside two standard deviations, we have 95% certainty that the observed difference is not due to chance aka there is a 95% probability that you are guilty .

[@syamilysasi8677]

@@TenorSaxx wow! Thank you so much! I didn't expect this much elaborate explanation.. you are very generous!! Thank you..!!

[@TenorSaxx]

@@syamilysasi8677 If you understand, I am happy with any effort!

[@ShubhamKapoor]

True, the video flipped the Null and alternative, I was so confused when I was going through the video first time.

[@nurjahanblaskar6105]

Alternative hypothesis is what a researcher has to prove and the null hypothesis is what other results could be

[@MicroLiberman]

Nice! Let say we do not want to say ACCEPT the Null hypothesis because there may be an error. So we say, we FAILED TO REJECT the null hypothesis.

[@edanmacuana6434]

confused... coz if i am not mistaken, null hypothesis use the symbol of "equal sign" and alternative hypothesis use "symbol of inequalities" like , and is not equal to.

[@tausifmulla]

Exactly.

[@bhavanisnvl5509]

True.

[@harineepriya]

Yes

[@dineshlakshitha7309]

Actualy wonderful explanation 🙂💪💪

[@esmatsahebdil7499]

Nice video. Thanks for making it simple and understandable.

[@humeraishaq6744]

I want to ask one question, shouldn't the null hypothesis be only with equal to sign. As i have read in many books that NUll hypothesis is A=B, where as alternative hypothesis could be A not equal to B for two tailed and A>B or A

[@wooddog007]

@3:00 ... i think the H0 and H1 are backwards. H0 should be 125k. Your friend thinks it is more than 125k, so you would reject the null hypo that it is equal to or less than 125k and conclude that your friend is correct.

[@myeverydaylife]

Most statistical texts would argue that the null hypothesis must contain some form of equality which contradicts your second example. P.S. I'm a data scientist by profession so I would appreciate you shedding some light on this.

[@andreamireles8508]

Loved this! Thanks for posting!

[@syamilysasi8677]

Good presentation! Congrats👍

[@tuannguyen4802]

Amazing explanation !!

[@dfig1020]

thank you very much!

[@rjsaloni8427]

Thank you soooo much!! Lots of love from india 😄

[@popohammers9095]

That's pretty useful. I think this video helped me. Ho: This video helped people understand the two types of hypothesis. Ha: This video did not help understand the two types of hypothesis.

[@Nazeerul_Hazard]

I don't accept this. But I don't reject this either :)

[@RAVITEJA-mf1vf]

thank uu soo much.. helped me a lot...🔊

[@mosama22]

Thank you for the beautiful video :-) :-)

[@Joyboy0101]

@365 Data Science I'm a little confused as to which hypothesis is supposed to be the counter and which we're trying to prove, since I watched a video by Investopedia where they said that the alternative hypothesis is ultimately the one we're trying to prove, while the null hypothesis is the one that's true until proven false. But my professor does the opposite. Does it really matter which hypothesis i set to = and =/=?

[@motuns12]

am same do not know which are are meant to prove

[@aang7505]

great explanation!

[@alecdelu83]

I really liked the video! thanks :)

[@jomarkfrancisvelasco8286]

Example 2. The null must be, mean is equal to $125000. while the alternative is the mean is greater than $125000.

[@Prosperous903]

fantastic, thank you so much

[@MrGuyTsurStiGy]

Awesome one! I am here for more!

[@teacherlaiarcenas4817]

thank you so much...

[@MiahB513]

Thank you for the video, very helpful!

[@kingsleykenechukwu3472]

Let’s get out of politics and into hypothesis 🤧, nice lines. Multi talented I see

[@bigguigomis7585]

Thank you!

[@devireddybrunda2955]

well explained

[@luisfernandocuestasanchez4343]

does the null hypothesis has to always be associated to an equality ( equal, greater than or equal, less than or equal) ??? I saw that you gave a different example with the salaries

[@humeraishaq6744]

Kindly help me with the concept of degree of freedom. Not simply n-1 or f-1. What is the basic concept of it. In your style. Please

[@juzores1]

Still complicated and unclear ,but thank you for trying to help .

[@carlosmspk]

What is it you don't understand? Perhaps I/someone can help

[@MohamedMostafa-kg6gk]

Thank u

[@alyafey69]

this was sooooooooo helpful thanks a million

[@virenbachani5305]

Wonderfully Explained!

[@hmmmm8105]

Thank you Kanye very cool!

[@TheSmartShowChannel]

I was taught that the alternative never has = sign in it

[@franciscopugliese1981]

same

[@ChandraKiran]

yes... me too... now what should i believe??

[@woozhenchee2886]

@@ChandraKiran two-tailed tests is denoted by an = or ≠, one-tailed test is denoted by >, ≥ or

[@qasimsultan3843]

Can any one solve this problem plz Q#1: a) State the null and alternative hypothesis for a test to determine the validity of the claim b) Identify left, right, or two-tailed test type The 2011 average service time for Fast Fatz Burgers was 56 seconds. The store manager claims that the 2012 average is more than 5 seconds faster. Speedy Solutions is a delivery company that claims all deliveries average less than 72 hours. At least 60% of Indians vote in presidential elections. Chance of developing heart attack is 40% for women.

[@samvoor496]

very useful, thank you

[@jdkanchana5289]

do u guys have videos in normalization?log transformation?unit vector normalization?standardization vs normalization.some practical examples of these. z transformation?lasso regression? fuzzy rule based induction? aprior algorithm?

[@festivetoe9492]

This was awesome thank you so much.

[@Sam-tg4ii]

3:15 I think this is incorrect. The null hypothesis should contain equality (

[@cryofreeze9815]

you are correct

[@rm9994]

Why is there a zero by the mean sign. What does it signify? Does it need to be there? It is in both null and alternative hypothesis

[@NehaKumari-wh8qn]

Thankyou❤️❤️

[@pinkudas2670]

Hi! Do we frame a hypothesis in qualitative research???

[@Thewiseenigma]

Sir Actually my question is that after getting p value i have i have to reject null Hypothesis ok i reject it but question is asking me is a is worse than b or a is bater than b so I'm unable to decide between them. Due to covid 19 I can't go college or somewhere else to purchase book and Im doing a stastic course but the teacher isn't responding please tell me the ans

[@manoj238]

Thank you.

[@sonnygad7008]

GOOD JOB

[@RevoOnRev1337]

before the 1 minute part: i slepp after the 1 minute part: w o k e

[@ajble]

wrong hypothesis testing set on 3.19. you have to contain equality sign on H0 always

[@narafied]

i dont understand :( why isnt the null hypothesis that the avg salary is 113,00 and the alternate hypothesis that it's >113,000 (so one tailed test) since paul thinks that the avg salary is greater (125,000) than the actual reported mean (113,000)?

[@Ojanto226]

in some places it will be like - Guilty until proven innocent.

[@TheStatsFilesDawnWrightPhD]

I'm pretty sure you made a mistake in your second example. The null must always be a form of equality and cannot be a >.

[@TheStatsFilesDawnWrightPhD]

Forgot to say this is an otherwise nice video.

[@365DataScience]

Hi! :) It is true that generally in a one-sided test the equality sign is often included in the null hypothesis. Moreover, we have not seen a conclusive source stating that one or the other is correct (or more correct). But let’s assume it does make a difference. We know from probability theory that for a continuous distribution the probability of obtaining a single number is actually 0. So, we can step on that property, or P(X=x) = 0, and we have the consequence: P(X>=x) = P(X>x). In other words, the two are equivalent! And there can be an argument about it, mathematically it makes no difference. Best, The 365 Team

[@TheStatsFilesDawnWrightPhD]

@@365DataScience Thanks for your response but you are missing the thrust of my comment. The null must always be a form of equality. It does not have to be an exact equality =; it can be =. There are many published/peer-reviewed sources I could cite on that point. The null is always the "no difference" or "no change" or "nothing going on here" hypothesis. We assume it to be true and use the sample statistics to try to prove it false. But we can never prove the null to be true, so we should never say we "accept the null." Rather we should only say we "fail to reject" the null. Granted in the real world, people do "accept" a null that is not rejected, but that is theoretically wrong. Again, many, many published sources could be cited on that point as well. That is analogous to the criminal trial example where we say the defendant is "not guilty" rather than "innocent" when we fail to reject the null that the defendant is innocent. My concern is that intro stats students will be confused by your terminology and presentation because almost all intro stats books say just the opposite of what you say in Example 2. Further, the only hypothesis you can prove is the alternative, so you should make it your test hypothesis, e.g. Paul's claim should be the alternative, Ha: the mean salary is > $125,000. That would make the null, Ho: the mean salary is which points to the right tail. If that probability is smaller than our significance level, we would reject the null and accept Paul’s claim the mean salary is greater than $125,000.

[@sungenik5226]

Thank you for your clarification. I took this to my Professor and he had similar sentiments to what you said.

[@TheStatsFilesDawnWrightPhD]

@@sungenik5226 Keep making your videos. We all learn best through mistakes. I learned just by double-checking what I thought about your presentation. Thanks.

[@TheDigitChannel]

INNNovation!

[@benitokestelman4545]

Not clear why it's called a null hypothesis vs. just a hypothesis. Also, the researcher is not "trying to reject the null hypothesis" - he is trying to determine whether to reject it or not.

[@anujamishra8801]

Alternate hypothesis cannot have equal sign. 3.23

[@hemanthm6015]

Who is born to Reject - Null Hypothesis 😂

[@rthmjohn]

I'm sorry but example 2.0 is entirely wrong. Your null hypothesis would be Glassdoor's currently accepted value of the median salary. The alternative hypothesis would be the claim that the median salary is higher. The easy way to remember null vs alternative is that the null hypothesis always has some statement of equality in it.

[@SketchyHDpro]

the status q-Woah

[@Stufwgbkuvsqfvuivdwg]

your voice is cool

[@gueklantan8461]

Why it call null!!!!!

[@maheshvenkat9956]

Equality should always be in Null Hypothesis.Hence you are wrong.

[@zephirothia]

Researcher would only try to prove a null hypothesis wrong? Never to prove it right? Sorry for the silly question

[@alexlenart1873]

If you cannot prove it wrong then it is assumed to be correct

[@akaashik]

Isn't this so obvious when you want to test something? Do we really had to give it some complicated naming and make it really sound complicated? Welcome to data science.

[@MrLoutee]

basic concepts in programming

[@MisterBinx]

So we are only trying to falsify the null hypothesis. If we can't then we accept the null hypothesis.

[@nicholaspayne2598]

Why is this a hypothesis testing problem?

[@wunderwaffeyt4077]

1:07 that did not age well

[@MelakuKK]

teleyalsh

[@wildzach]

Why can't statistics just not exist

[@Wjdveiwlslfbeb]

WOW

[@malakabdullah8316]

watching on the elections day :)

[@richardong4757]

So... Can the null hypothesis be "the mean data salary is less than 125,000" and the alternative hypothesis "the mean data salary is more than or equal to 125,000"???

[@lloydedersonquiambao9345]

I guess research is not a subject for me😂

[@HamnaAzhar7]

3:27 but i read somewhere that null hypothesis always contains equal sign...and that's the thing that confuses me in every single question

[@aleksikulmala6397]

Thats only the case for two tailed tests. One tailed tests can be greater than and equal to, or less than and equal to.

[@HamnaAzhar7]

Jack Reacher no i meant that it said the hypothesis that contains both signs will always be considered null hypothesis and the one that contains inequality will always be alternative hypothesis

[@HamnaAzhar7]

Jack Reacher so according to them the second one would be null hypothesis and the first one the alternative hypothesis

[@leelacabin]

2:06

[@Stein_NZ]

I think you should make another video because this is incorrect.

[@timthefish2626]

Is curious droid narrating this?

[@jaiswalv4384]

Innocent until proven guilty

[@zilongmetei9243]

Ok let's get out of dirty politics

[@RevoOnRev1337]

ok