Tutorial 32- All About P Value,T test,Chi Square Test, Anova Test and When to Use What?
HTML-код
- Опубликовано: 21 авг 2024
- Please join as a member in my channel to get additional benefits like materials in Data Science, live streaming for Members and many more
/ @krishnaik06
Please do subscribe my other channel too
/ @krishnaikhindi
----------------------------------------------------------------------------------------------------------------------------------------------------------------
Many people has asked me about various products and tech gadgets that I use for recording purpose. Here I have put up all the products. Definitely check it out :)
💻Primary Laptop : amzn.to/3AuKL8z
💻Secondary Laptop: amzn.to/3wrznrb
🎧 Headphones for streaming: amzn.to/3hdbzCS
🎧 Wireless Earbuds: amzn.to/3wg6aiD
💺 Chair for Studio: amzn.to/2TuJ0Ia
🖥️ Monitors : amzn.to/3AvQkD....
📷 HD Webcam: amzn.to/3hvqBmm
📋 Pen Tablet For Writing: amzn.to/3xdkGsN
🧸 Soft-Toy: amzn.to/3hDUyk8
📷 DSLR Camera: amzn.to/3dKnsxR
Connect with me here:
Twitter: / krishnaik06
Facebook: / krishnaik06
instagram: / krishnaik06
Amazing, its like my 5-6 hour online class video merged into a 12 minute video.
Good job !! Some parts of the explanation can be improved, especially your point about ANOVA test when a categorical variable has more than 2 possible values. Consider slowly down and collecting your thoughts together and your videos will be even more effective.
Yeah, I have the question, when he takes Gender and Age Group, then he used Chi-Square test, but later said when a category containing multiple values (not binary) then we use ANOVA.
Sir I wish to watch all your vedios ..I subscribed it.. pl send all liks regarding Excell,data types, hypothesis testing,
Typically you reject Null Hypothesis or You Fail to reject Null Hypothesis. "Accepting" H0 or Ha term is typically not used..
I had the same point, either we reject Null hypothesis or we fail to reject it.
Take null - *they are independent*and then proceed.
Exactly. You don't accept either alternate or null hypothesis.
This was a good overview of the different hypothesis tests. Looking forward to seeing more videos from you in this series. 😊
Thanks SAR
You cannot take up any test, like if you want to use a binomial test, then your question should follow that binomial distribution.
This video clear 80% of our Hypothesis testing concepts. It's a very good explanation.
What about the remaining 20% of the concepts
@@sanjeetsingh-iz1rb significance level is 20% in this case
Explanation was very good. I would like to know if my assumptions mentioned below are valid. Hope you acknowledge this.
1. select k best can be applied on both classification and regression problems
2. T-Test can be applied on a categorical feature which has only 2 distinct categories and when sample size is < 30
3. Z-Test is same as T-Test but is applied when sample size > 30
4. ANOVA Test is applied to categorical feature which has more than 2 distinct categories
5. T-test, Z-Test & ANOVA tests are applied only when target has continuous values .
I.e, when we are working on regression model
6. Pierson Co-relation Co-eff can be applied only on numerical features. It can be applied between a feature & target and also between features
If we find 2 features that are not co-related, we can remove one of them.
7. Co-relation matrix can be applied only on numerical features
8. Chi sqr test can be applied only on categorical features
2. T-test applied on one or 2 numerical features.
t-test and ANOVA work on numerical and continuous values.. yet in classification, we are using dummies the dependent feature(target column). Hence it can be applied.
@@mooventhc1686 Thanks much. Correct me again please. T-Test, Z-Test & Anova-Test are used when our target column is having continuous values. I agree. But what should be the type of input feature ? Categorical / Numerical ? On which input feature type T test and ANOVA tests are applied ? Thanks in advance
Watching you hustle...i push my limits 🙏
Thanks you so much Sir.
Hey its three year's, what were you hustling, did you achieve that?
I think the starting point of Data Science is the Analysis of Data and these tests determine the Algorithm and the Regularization method to implement to minimize the cost function (RSS).
Read recently that
1) Co-variance and Multi-Collinearity would have impact on the Coefficients and NO impact on predictions
2) There are L1 and L2 Norm regularization methods. A study (Mark Schmidt CS542B Project Report December 2005) says that L1 with Optimizing Least Squares is better than L2. Reason being that L2 does not address Parsimony (sparsity) of the model and Interpretability of the coefficients values and all it aims is Shrinking the Coefficients. L1 regularization has many benefits of the L2 and yet, sparsity and interpreting coefficients is easy.
While above two are understandable in English but not as Statistics. May I request you to cover these, if possible, in your next session.
Its so nice to see "whys" and "whens" in this video, which I think is the matter for Data Scientist. Great Work Krish. Please keep it going with more Whys and Whens.
One category Feature --> One sample proportional test or Z test
Two Category Features --> Chi-Squared Test
One Continoues feature --> T -test
Two continoues variable --> Co-realtion plus t test
numberical plus category variables--> annova
I have watch more than 5 videos and still could not understand and finally sir videos has made it so comprehensible.....watching this video just 2 hrs before my exam😂
Amazing video sir...
It has cleared my doubt on one of contradictory topic.
Thank you very much for this teaching........
1. Test on : One categorical feature - Two subclass : One Sample Proportion Test
2. Test on : Two categorical features : Chi-Square Test
3. Test on : One continous(Numerical) feature : T-Test
4. Test on : Two continous features : T-Test (Correlation used here)
5. Test on : One continous feature and One categorical feature with only Two subclass : T Test
6. Test on : One continous feature and One categorical feature with More than Two subclass : ANOVA Test
In All Cases, Reject H0 if p_value(significance value)
What about z-test...when the sample size is greater than 30?
It comes under similar categories as that of T test and used when sample size is large
The p-value is the likelihood of the observed data, given that the null hypothesis is true. The more it is low, the more we are confident to reject H0
Q- why we use P=0.05 or 5%?
A- From experience or we can say from previous experiments we have concluded that from a population about 5% outcome is defective or we can say we have to reject that amount of data that falls within or equal to 5%.
You're doing a great job, sir. Understanding these concepts is as important as knowing how to code.
Thank you so much Sir...now i learned and understand the difference in between the T test, correlation, ANOVA.. P value significance ...etc
You need a correction: Rejecting the null hypothesis does not mean that we accept the alternate hypothesis.
We never accept the alternate hypothesis. We only reject the numm hypothesis or fail to reject. We don't do anything with the alternate hypothesis.
could you point to some more references of what you have said, cause till now even i thought that if we reject H0 we accept H1, if not references then maybe explain a bit more as to why. thank you!
best channel for learning statistics i've found so far. Great job
Annova test--
when we have one numerical variable and categorical variable
where categorical variable has more than two categories
T-test--
when we have one categorical variable and one numerical variable
where categorical variable has only two categories or one continuous variable
one sample praportional test--
when we want to campare values from only one sample(sample is categorical)
chi-square test--
when we have two categorical sample
correlation test--
when we have two numerical variable
best playlist i have seen ever
We can only reject null hypothesis but never accept alternate hypothesis. Based on test we can only conclude that we either have evidence in favor of null hypothesis or not.
Great explanation, much better than the education I received in the last three months combined.
You speak very fast! thank you for explaining so well
play video on 0.5x
p-value
Given a chance model that embodies the null hypothesis, the p-value is the probability
of obtaining results as unusual or extreme as the observed results.
Alpha
The probability threshold of “unusualness” that chance results must surpass for
actual outcomes to be deemed statistically significant.
THIS IS YOUR BEST VIDEO SO FAR !
It's the best overview of tests I have seen on RUclips.
Awesome dear sir.... Thank you.
Hi. I don't know much details about the different hypothesis tests but I have learned in my probability class that if the correlation between two variables is zero never say that the two variables are independent but if the two variables are independent, then the correlation must be zero. So How could you apply the correlation test to find the dependency? It will be really helpful if you explain further. However, Thank you so much for your dedication to providing these videos free of cost.
Correlation is the test done to check if two variables are related or not, if yes then how strong is the relationship. We do something called Hypothesis testing in order to check if the relation shown by the variable for it's respective sample(which we used to compute the correlation) is significant for the population data too. By doing this test we come to know if the correlation shown by the variables are significant or caused by chance or due to sampling error
I am grateful for the brief information for the various test in the hypo & null hypo. helpful
Super krish naik jeee crystal clear explanation …..preparing for PhD
It’s helping me a lot thank you once again
Thank you very much, Krish. Tomorrow I have a mock interview on Machine Learning. a lot of thanks to you.
Which company ??
Oh my god Krish got angry 7:02😂😂😂,jokes apart you are gr8 teacher.
sir! really blessed to watch your videos!! ur passion towords it make me feel enlightned 💯🙏
Very nice video.
Learning points of this video:
1. Test on : One continues features , Hypothesis on : mean , Comes under : One Sample Test , Name of Test : One Sample T-Test , Accept & Rejection hypothesis criteria on what scale comparison : p value
2. Test on : One categorical features - Two subclass , Hypothesis on : proportion between two class , Comes under : One Sample Test , Name of Test : One Sample Proportion Test , Accept & Rejection hypothesis criteria on what scale comparison : p value
3. Test on : Two continues features , Hypothesis on : correlation , Comes under : Two Sample Test , Name of Test : Correlation with T-Test , Accept & Rejection hypothesis criteria on what scale comparisons : correlation & p value
4. Test on : Two categorical features , Hypothesis on : proportion between two class based on other class , Comes under : Two Sample Test , Name of Test : Chi-Square Test , Accept & Rejection hypothesis criteria on what scale comparison : p value
5. Test on : One categorical feature - Two subclass & One continues feature , Hypothesis on : Difference of mean between two class(variance) , Comes under : Two Sample Test , Name of Test : Two Sample T-Test , Accept & Rejection hypothesis criteria on what scale comparison : p value
6. Test on : One categorical feature - More than two subclass & One continues feature , Hypothesis on : Difference of mean between more than two class(variance) , Comes under : Two Sample Test , Name of Test : ANOVA , Accept & Rejection hypothesis criteria on what scale comparison : p value
Super and Great, This was what I was waiting for long time, Thank you again 🙏
Krish we understand the concept but don't know how to implement it in real dataset on python or R please make video on that by doing in jupyter notebook or rstudio.
4:18 📖
5:56 📖
7:09 📖
8:51 📖
Thank you so much for putting it all together in this concise video.
love real whiteboard lessons like yours..... my professors are dull and just run powerpoints during lectures half asleep.
Excellent teaching
Thank you so much. I love your method and pace of teaching.
Conclusion:
One category Feature --> One sample proportional test or Z test
Two Category Features --> Chi-Squared Test
One Continoues feature --> T -test
Two continoues variable --> Co-realtion plus t test
numberical plus category variables--> annova
Your explanation creating interest to learn statistics
t-test is actually more suitable for comparison of two populations samples. Analysis of variance (ANOVA) is a statistical technique that is used to check if the means of two or more groups are significantly different from each other by analyzing comparisons of variance estimates. But Krish mentioned in this video, t-test can be used when you have one numerical variable at around 8:20[ 8 minute, 20 seconds ], is it true statement or is it just mistake in flow? Thank you so much for investing your personal time in advancing the common good in our community, God bless you.
i think you are right.
The t-test is a statistical test used to determine if there is a significant difference between the mean Value of two groups. It is commonly used when working with """numerical or continuous """ variables.
ANOVA is typically used when working with numerical or continuous variables.if there are significant differences between the means of three or more groups.
t test is applied if there is one categorical and one numerical value but here you told that for t test only one numerical value is seen.
Best explanation.. 👍👍
this guy came 4 years too late for me! thanks for this
P is low
Null will go
P is high
Null will fly
Thank You sir... It was very knowledge full
krish, I have observed that you mentioned to use T - test for two numerical variables and again you mentioned correlation test.
Wonderful explanation, thank you very much for making it so easy and interesting
Amazing this has given me a clear understanding.
A candidate sat in recruitment tests for the job of data scientists by one of the top leading firms of US. He was confused whether the salary in the company is good or not. So, what he did was that he took a survey of 14 employees working there. Their salaries would be given as input and candidates would like to test the hypothesis whether there is no significant mean difference in salary of data scientists given input mean. We have to return True if we can decline the hypothesis else False (Take threshold of 0.05) plz solve this question
Crisp and to the point. Good one Krish.
"Eldaly" Haha, made my day
Thanks for the compacted video & all the tests at one place. I don't think so there is any other video on you tube explaining all the tests in such short & meaningful way. Nice video.
Also, just got a doubt what test do we need when there is a categorical & numeric variable combination?
If I'm not mistaken, acc to what he say if there are combination of categorical and numerical where both categorical and numerical variables has more than two distinct sets of value or group then Anova test should be apply.
we can't say that the t-test and chi^2 is used only for the categorical variable. We can use it or analysis of mean, variance etc.
Very informative video!!😃
Sir you explained it very well, in a very easy to understand way. The only problem was audio quality. Else everything was perfect.
Thanks for the lucid explanation.
thank you for doing this video. it is a very useful and good explanation with a simple example.
Very well explained Krish
Excellent Teaching. Thanks
Very good video again as earlier. The way of connecting different concepts together is the difficult part for beginners and students. Your approach to answering the above issues are excellent Krish. Thank you very much. Please continue your good job for this world.
Its great to seea good video on hypothesis testing.... good going..
I expected the series as a playlist,
I directly reached this video from you tube, and I dont know what to do with this series next
No i button, no playlist, I wont go to your page @krish and search for the series now, you need your content to be more accessible !
maybe an intro to the series before every video, quick 10 seconds..so that its easy to locate, for someone who would want to go to your page
Excellent video, describe concept clearly
Helpful explanation.
Thank you for your effort sire
thank you so much!! you make things easier!!
big thanks Krish!
MaNY Thanks bro , you sorted me out!!
Thank u so much sir it really helped me a lot to understand this concept
Thanks a lot. Thanks for excellent explaination
Hi Krish, thanks for this amazing video. Could you explain this using python with the sample data set.
Indias own Andrew NG is krish naik ji.... Rather krish NG
Please enhance the audio quality rest things are very nice and informative
Excellent Class Sir ....
Thanks Krish
Very good explaining sir. Thank u ❤
You sir are amazing! Thank you for this video!
watching the video for second time for revision. Thanks
Thank you so much for this nice explanation
Simply superb
Hello Krish/ Subscribes of Krish ,
Kris mentioned T test can be used only for continuous variable in the beginning , but in the end at 11:11 Krish is saying it can be used when we have one continuous variable and one categorical variable( i.e only 2 categories), please reply me i am confused after this krish.
It's about paired and unpaired t test
Awesome explanation thank you
Great video!
Hello,
Can you please add a video implementing the pipelining technique for ensembling more than two different algorithms together.
Thank you for the amazing explanation sir. If we are comparing continuous variables we use corelation, wt if we are comparing 2 continuous variables in 2 groups as in comparison of newborn anthropometry (length and wt) in anemic and non anemic group. Kindly guide me. Thank you
very nicely explained. Thank you
Very nicely explained
Easy to understand.. You have enlightened me :D
Thanks Krishh for the awesome video.
Thank You, very clear explanation
You teach awesome sir
God Bless you Sir ...
Very helpful thanks
superb well explained appreciated
wonderful explanation
Well explained!
Thanks
Excellent tutorial
when it comes to correlation, do we have to use T test again?
We simply see the correlation value between two variables. Close to 0 means no correlation.