Great question, and in fact you're totally right about using a binomial test, it is even a little bit better statistically speaking, since you exactly know the 'reference population' numbers (33% in this case). I will say that in practice, if you're right on the significance limit of a test you're in murky water anyway. With these triangle tests, you assume that if your new product is actually different, that one will be chosen more often than by pure change. So for some extra comparison, if you use the setup discussed in the video (1 out of 3 is the new one, so 33% random chance), you'd need 3 out of 5; 6 of 10; 10 of 20; or 22 out of 50 samples to say that this new product is significantly different from the old one. As you see, if you take about 20-50 samples, you're at that same roughly 46% that an average Chi square gets to (and the chi squared numbers will also chance when you take fewer or way more samples). But again - thanks for this question, this is how we all learn to do better Six Sigma. And just to be overly complete, the answer to your second question is that binomial is preferred when you accurately know the expected distribution ahead of time (like how often a coin should land heads up), while chi squared is preferred when using the data itself to help assess the reference distribution and there could be some sampling error in that reference too (like comparing how many defects we produced last year to how many we had this month - since your data from last year isn't "the truth" but also has sampling errors).
Great explanation..Thanku for explaining it so well. I have one more question, gor example I am doing a triangle test where I have two control sample which contains 100% salt and one test sample which contains 90% salt. and I am asking consumers among three samples, which tastes different. So here should I use binomial or chi square and why?
@@oshinsahni1742 again, both will work fine. The statistically preferred option will be the binomial, because you know that the prediction value is 33% (you’re not comparing 100% to 90% salt, but the assumption that your assessors will pick the new product 33% of the time randomly if they can’t actually sense a difference between the two products)
Interesting video and well explained, thank you! Some sensory text books recommend to let the panelists test all different combinations (AAB, ABA, BAA, BBA, BAB, ABB). How would you analyze the data? Just by assessing all responses independently in a chi2-test?
You don’t have to split all responses independently. In fact, statistically AAB and ABA are the same, for instance. What you do in this case is to use the total combined distribution of all offerings into 1 percentage: if you gave 6 people each one of these variants, you test for an expected 50%/50%; if you gave 10 AAB, 10 BAA and 10 BAB, in total you handed out 50 A’s and 40 B’s, so chi test with expected 55,5%/44,5%.
Not really, it is always better of course try and keep the test conditions the same if possible, but actually this triangle test is particularly suited to be used in 'market conditions'. Do keep in mind that it is a very suitable test for food products, packaging and other things that you test with human sensory input. If you've got specialised measuring equipment to generate a value per product, using a t-test is almost always better. The triangle test is used when you want to assess whether or not humans can spot/taste/sense a difference between two versions of a product.
![I.N "HALLUCINATION" | [Stray Kids : SKZ-PLAYER]](http://i.ytimg.com/vi/n5B5q1Hwt_U/mqdefault.jpg)
Well and clear explained! Makes my school assignments a lot easier to understand :)
Thanks for the compliment and happy that my videos can help you 😎
Can we use binomial testing also for triangle testor chi square is better. How do you choose?
Great question, and in fact you're totally right about using a binomial test, it is even a little bit better statistically speaking, since you exactly know the 'reference population' numbers (33% in this case). I will say that in practice, if you're right on the significance limit of a test you're in murky water anyway.
With these triangle tests, you assume that if your new product is actually different, that one will be chosen more often than by pure change. So for some extra comparison, if you use the setup discussed in the video (1 out of 3 is the new one, so 33% random chance), you'd need 3 out of 5; 6 of 10; 10 of 20; or 22 out of 50 samples to say that this new product is significantly different from the old one. As you see, if you take about 20-50 samples, you're at that same roughly 46% that an average Chi square gets to (and the chi squared numbers will also chance when you take fewer or way more samples).
But again - thanks for this question, this is how we all learn to do better Six Sigma.
And just to be overly complete, the answer to your second question is that binomial is preferred when you accurately know the expected distribution ahead of time (like how often a coin should land heads up), while chi squared is preferred when using the data itself to help assess the reference distribution and there could be some sampling error in that reference too (like comparing how many defects we produced last year to how many we had this month - since your data from last year isn't "the truth" but also has sampling errors).
Great explanation..Thanku for explaining it so well. I have one more question, gor example I am doing a triangle test where I have two control sample which contains 100% salt and one test sample which contains 90% salt. and I am asking consumers among three samples, which tastes different. So here should I use binomial or chi square and why?
@@oshinsahni1742 again, both will work fine. The statistically preferred option will be the binomial, because you know that the prediction value is 33% (you’re not comparing 100% to 90% salt, but the assumption that your assessors will pick the new product 33% of the time randomly if they can’t actually sense a difference between the two products)
Thanku so much for explaining it so well. It makes sense to me now.
@@oshinsahni1742 my pleasure!
Interesting video and well explained, thank you! Some sensory text books recommend to let the panelists test all different combinations (AAB, ABA, BAA, BBA, BAB, ABB). How would you analyze the data? Just by assessing all responses independently in a chi2-test?
You don’t have to split all responses independently. In fact, statistically AAB and ABA are the same, for instance.
What you do in this case is to use the total combined distribution of all offerings into 1 percentage: if you gave 6 people each one of these variants, you test for an expected 50%/50%; if you gave 10 AAB, 10 BAA and 10 BAB, in total you handed out 50 A’s and 40 B’s, so chi test with expected 55,5%/44,5%.
Thank you sir, well explained. Is there any particular testing condition required for the triangle test, like temperature?
Not really, it is always better of course try and keep the test conditions the same if possible, but actually this triangle test is particularly suited to be used in 'market conditions'.
Do keep in mind that it is a very suitable test for food products, packaging and other things that you test with human sensory input. If you've got specialised measuring equipment to generate a value per product, using a t-test is almost always better. The triangle test is used when you want to assess whether or not humans can spot/taste/sense a difference between two versions of a product.
@@TomMentink how do you code and decode samples?
Thanks for this ..... great learning 😎
Thanks for the compliment - I'm glad you found it useful!
If there are any similar topics you'd like me to cover, don't hesitate to ask 😉
Hi sir
Hi yourself 😃