We do not use a t-test for a one-sample test for proportion. We use the z distribution. I teach undergraduate and graduate statistics and have reviewed a dozen statistics textbooks and with 99% confidence, claim that we do not use the Student's t-distribution when testing a proportion. From the Stats Stackexchange: "The reason you can use a z-test with proportion data is because the standard deviation of a proportion is a function of the proportion itself. Thus, once you have estimated the proportion in your sample, you don't have an extra source of uncertainty that you have to take into account." Z distribution doesn't ask for sample size to determine the critical value, whereas t distribution does. However, as sample size gets large, z and t converge until t = z. For example, the sample size of 1000 would have no difference between z and t values.
The z-test for proportions is DERIVED from the binomial distribution under the assumption of a large sample size. The key idea behind this test is based on the Central Limit Theorem, which states that the distribution of the sample mean of a large enough sample will be approximately normally distributed, regardless of the shape of the underlying population distribution. hence you use z-test with proportion data
I took the data analytics course on Coursera and they also taught to use only a z-test for proportions. Rushed to the comments for confirmation so thank you lol
I am from the university of Jos, Plateau State, Nigeria and l have never learned well during the masters degree statistics classes l took for one year+ but you make it look simple. You are a good communicator and l feel more confident about my statistics knowledge because of you.
@@joed2444 ohh I see. I should add then that this video really only tells you things to memorize about these topics. If you dive deeper into the mathematics behind every statistical test, you will not need to remember what test to apply in which situation, it will just make sense. Just my two cents if in case you are interested in studying stats.
There is nothing wrong with the 1-sample z test. It produces the correct answers as long as you have the population parameters. But because we rarely have the population parameters we don't use it. Its not inherently flawed as implied by the host of the video. It's pretty easy to explain why it isn't used very often, like I just did. Otherwise, great video.
I was going to say the same thing. In addition, when the sample is greater than 30, both tests are pretty much the same. For the Chi-square, he should have said that each "bin" needs to have more than five observations. Two of them were below five.
@@Canuck1000do you know the difference of 2 sample t-test proportion vs chi squared? I feel they are quite interchangeable, like the depression example he mentioned. Can it be used for chi-squared as well?
@@claireli5044 The issue that is discussed here is the z-test vs t-test (population vs sample). The z-test is valid if we know the information about the entire population, but is very difficult to obtain as Z3r0 said above. It should not be automatically rejected. In the end, if n>30, both tests will give you the same results (if it is a sample still use the t-test though).
@@claireli5044 Hi, isn't the t-test for comparing the means of two or more different continuous variables and the chi-squared test for nominal and/or ordinal variables? If I am not mistaken I don't think they are interchangeable. I would appreciate it if someone can correct my understanding. Thank you.
I’m a mechanical engineer. Of course we took the appropriate statistics and probability class but never saw the real significance until my 5th year of being an engineer and working at GE healthcare. Now I apply the DMAIC approach as much as needed.
You are damn right, buddy! I had such a similar experience, and I believe to many out there! I found that the DMAIC thing of LSS is so interesting. So, you are practising it within the healthcare industry? Wow, that is great!. Maybe we can catch up for more experience sharing...😊
Simply the greatest explanation. I spent hours trying to get the main gist and difference of all these confusing test. This video was the brilliant saver for me. Thank you so much !!!!!
You use the Z-test when you know something about the population standard deviation. This doesn't happen very often, so the t-test is more common. Saying you should never use it, or that it's "bad" or "dumb" or "unprofessional" is just NOT accurate. It's just more RARELY used. Throwing it out shows a lack of understanding.
I’m not statistician but I’ve been in the statistics classes at least all three or four you might take during your undergraduate year and when you learn about those particular test they also tell you when you should use them. That’s part of learning about the actual test. I think that’s where we’re having a breakdown in the education process now statistics is more college level but even on a more basic level we might learn how to do math but we don’t learn the principles behind the mathematics you should learn the principles behind the statistics tests then you know when to use them
Thank you very very much I have an exam tomorrow and you explained it clearly that I understand now I really appreciate your efforts you are a genius thank youuuuuuuuuuuuuuuuuuuuuu
Kody, Thank you so much. I have spent days trying to figure out ANOVA versus Chi Square versus T tests and so on, and you made it so easy. I am really grateful --and also a little baffled that so many other sources make it so complex.
Thanks so much for sharing the knowledge... for FREE! However, one of my statistics teachers, used to say to me to use t-test for a small sample size i.e., of less than 30 with unknown standard deviation of a population with a normal distribution property. That explanation still holds?
Thank you so much! I wish every statistics class started this way. Getting to look at the bigger picture first and then jumping into details is always a better way to learn things.
Ok, first, I love the way you present the information = terrific!!! 2nd soooo easy to understand; you speak in everyday language. thank you so much for this video, I got it!!
Hello, This lecture has helped me to understand using T and ANOVA Test on categorical variable. Previous my thought was T and ANOVA is used only for MEAN difference.
Hopefully, I can gain a better understanding of this topic. My 2nd take with Probability and Statistics for my undergraduate degree in Psychology (Science) via Online.
I think proportions are for categorical response or variables and not qualitative as said here, qualitative research itself its so much complicated with thematic or content analysis of codes and quotes, regarding Z and t test I think its about sample size that dictates which one to be used, I stand to be corrected if I am wrong
Z test for proportions T test for means Chi Squared Goodness of Fit for 1 sample with 1 variable to test if it is different for the population (ie test if the distribution of race is different in 1975 population to the 1980 sample) Chi Squared Independence for 1 sample with 2 variables to test if they are independent (ie test if chess and IQ scores are associated) Chi Squared Homogeneity for 2 samples with 1 variables to see if they have the same distribution (ie to test if men and women have the same distribution of living arrangements) T test for slope
Hi Sir, i need your help. From below info, what you understand. Can you explain to me, pls? Hypothesis i) There is a positive relationship between salary and employee retention - BETA VALUE (-0.379), Pearson Correlation (-0.289) Result : Accepted ii) There is a positive relationship between communication and employee retention - BETA Value (-0.159), Pearson Correlation (0.110), Result (Accepted) iii) There is a positive relationship between job satisfaction and employee retention which impact their decision to stay : BETA Value (-0.115), Pearson Correlation (-0.136), Result (Rejected)
Talk about learning Math 10 Statistics class Online, and then a Research Methods class again online. Now I'm good with online, but not online for harder classes.
When he talked about the 2-independent sample test for mean, 2-independent sample test for proportion, and paired sample test, are those all t-tests just with different null and alternative hypotheses? Or are there different ways to do those tests other than the t statistic?