I have a question for the group related to an EOC in the CFAI Curriculum. Can anyone assist? For reference, this is Volume 1, question number 11 on page 380 in the curriculum. Please read the below in quotes and the part A and part B below:
[question removed by moderator]
Additional Comments
I understand that you need to run the Breusch-Pagan test and look at the tables in the back to compare with the result of n*R2. However, the solution uses the significance level of 0.05 as opposed to 0.01. Based on the tables in the back, at 0.05 you would reject the null, but at 0.01 you would fail to reject, so obviously how you answer part B would depend on what level of significance you use.
Does anyone know why you would use 0.05 in the tables as indicated by the solution in the CFAI books instead of 0.01?
Sorry man, you must realized (and me, I already realized ) that the CFA books are not good. The authors must improve their logic and/or to write correctly CFA books.
If you want to learn statistics, buy a statistics book. It is worth 100 times more the CFA books. But if you want to become CFA charter holder, try to guest what the authors wrote (they wrote when they were on vacation in a beach, drunk beers, I suppose) and accept the fact that there are many “implicit” definitions/conventions.
People are here, in this forum, to help you. We are all trying to understand this “bible”, written by “gods”. You must believe, learn by heart. If you questions about the logic of “gods”, you fail.
I was crazy when I read sessions concerning mathematics. CFA books are good for general culture, but really not good for the logic.
It is dumb to think that a CFAI book will teach all the basics and fundamentals of a certain topic. This is why CFA program is directed for bachelor people at least. Also, it is not directed to all kind of professionals. If you have experience on marketing and you’ve never seen a regression before, well, start learning from the basics. I can tell you that the exam won’t be extremely deep in the concepts the books show, better try to catch why regressions are useful, what types are there, learn the calculations and interpretations.
You can use any alpha you want. It will depend on how strict you want to pick the solution. Let’s say that the most common alpha is 0.05 because remember that the lower the alpha (lower probability of type error 1), the higher the probability of type error 2. So a 0.05 looks a good balance.
On the exam, they will be really clear what assumptions to use. Don’t worry too much about this.
I criticize only quantitative sessions, I don’t want to talk about non-technical sessions (which I found interesting).
It is dangerous to learn quantitative tools in CFA books. In fact, if you understand clearly what they said, you are not good enough in math. Because if you were good enough, you must realize what they said is ambiguous. The books just make you think that you understand, but in fact you don’t.
Why the most common alpha is 0.05? Most common for whom? If 0.05 looks a good balance (between what and what?), it seems not necessary to have 0.01 or 0.1 in CFA table (at the end of the book 1), right?
Ok, you said “On the exam, they will be really clear what assumptions to use”, I hope so. But we are all talking about the books here. I may be agree with you about what you said on the exam , but your argument is completely invalid about the books.
I emphasize that I don’t criticize all CFA books, I just criticize quantitative sessions in CFA books. Other topics is usefully and interesting for me.
if you understand clearly what they said, you are not good enough at logical thinking. Because if you were good enough, you must realize what they said is ambiguous.
I agree pretty strongly with you that the CFA Institute books are pretty poor for learning statistics (in terms of coverage and explanations/accuracy).
This was really what I was getting at. If everything in the problem uses 0.01 as the level of significance, but in the response, they base the answer on 0.05, I didn’t know if I was overlooking something in the reading, if it was an an error in the book, etc…
When going through these problems and you get stumped or something doesn’t seem right, the first thing you ask yourself is, “If I see something like this on the exam, what would I do?” If what you are saying is that this will be clearly explained to me on the exam, that makes me feel better.
I think I have a pretty good handle on the concept, but I just expected their explanation to involve 0.01 and the fail to reject/reject conclusion to be based on the 0.01 level of significance.
You learn how to interpret regression calculations, how they are used and why in an (for me) an adequate level of deepness.
You want to become a statistician / mathematician. If you want to achieve this by taking L2 exam, then you are a fool.
I have read the curriculum for Quantitative Methods for L2 and found it very attractive. I have not compared it thoroughly with specialized books (despite I have some) because it was not the intention. I can say that those 4 chapters include about 200 definitions, formulas and procedures; perhaps 1 or 2 are not 100% accurate. Believe me, it won’t harm you if you just want to achieve point 1).
If you get confused by the way an author explains the concepts or how her redaction is, then you won’t be able to adapt your knowledge in the real world. By the time someone talks about the same concept but with different words you will be crashed.
About the alpha 0.05 or 0.01, again, as I said don’t worry too much if an EOC question was a little bit confusing because it does use different alpha than a previous question. Alpha level is a choose in the real life. For the exam, it will be clearly dictated what level to use.
I have done the both levels until now, didn’t find any ambiguous question or data. However, expect the CFAI online vignettes to be brave sometimes (in difficulty and in hidden info), it is just to train you.
Depending on how reliable the data is, what type of model you are using, you can assert more accurate conclusions or not. Using an alpha of 0.2 is very common in logit and probit models for example (probabilistic models), linear regression commonly uses 0.05 and also 0.1; however you can also use 0.01, depends on the objectives of the researcher. Just remember the Type error 1 and Type error 2 relationship when deciding.
The CFAI books are not intended to be scientific or specialized books, indeed they are made that way to be much more digestible. Believe me, I have in my hands a book specialized in advanced Econometrics, it is even in my main language (spanish) and I have big trouble understanding 50% of it. You know why? Because that book is intended to be a guide for Master or P.hD economists or other professionals that use econometrics. This book builds its concepts and axioms over other concepts and axioms the reader must handle in advance. If not, you are crashed like I do.
With my all due respect, and really I don’t want to hurt you, but your logical thinking can be improved. In fact, I have (and I choose) the third option :
“I want to become a statistician / mathematician, and I want to become also CFA charter holder. But I’m not fool.”
I gain my knowledge from non-technical topics in CFA books, I learn statistics, math,… in statistics books, I familiarize with technical topics in CFA books, I must guest what authors want if their questions are not accurate (no choice). But I keep in mind that what they wrote is toxic.
I, and many people in this forum, try to understand (or already understood) what the fools (CFA authors) wrote, but it doesn’t mean that we are fool.
I’m not saying that you are not good in math, but you may be not good in logical thinking. A normal high school pupil can show you the ambiguity in the above question.
And you must distinguish “question containing sophistical mathematical concepts” with “question with missing data”. An example for the first one is “Prove that x^n +y^n=z^n have no solution in integers greater than 2” (Fermat’s theorem) and for the second one “Give my the age of my first neighbor if you know : my first neighbor’s age + my second neighbor’s age = 60”.
The first question is really difficult but it can be solved with certainty. The second one is extreme difficult for you, because you don’t know the age of my second neighbor. Me and my neighbors, we can answer it easily. You must guest the answer.
I hope you understand what I meant. If not, you can’t go further with the book in your hand.
The CFAI books are not intended to be scientific or specialized books, I agree. But it doesn’t mean that CFAI books must be ambiguously written to be much more digestible (digestible for you, maybe, but not for me).
I think you are making big trouble from what it seems a typo. I have said that the real exam won’t show those kind of ambiguities, nor mistakes. Is much easier to correct typos in a 40 pages exam than in a 3,000 pages curriculum. Some errata will appear inevitable.
I don’t get what you mean with “logical thinking”. My thinking is clear and I have elaborated. You are, in the other hand, stuck in this little typo problem.
I don’t really understand what you meant. But if you told about my English. I’m aldready conscient about it because English is my third (and second foreign) language. But I know I can improve it. But writing English is not what we are discussing here.
About logical thinking, maybe my term is not correct, but I can give you an example: If you can’t distinguish “question containing sophistical mathematical concepts” with “question with missing data” I described above, you are in trouble.
Buddy, I have never mentioned nothing about your English proficiency, set it clear…
I do differentiate those, but in this specific case the question didn’t describe any of those problems. The problem was on the solution.
Again, be safe from this happening in the real exam. CFAI has a policy for ambiguous questions, typos, mistakes, etc which always favor the candidates.
that because I read rapidly your reply in my phone.
About this question," A. Perform a statistical test to determine if conditional heteroskedasticity is present." which alpha must be chosen? If I choose alpha = 0.9124 (or a random number between 0 and 1) for instant, my answer is acceptable?
It is not problem of typo. It is problem of logic.
You said on the exam, there are not logical problem (or typo problem). I hope you are right.
And if you read topics in this forum, you will find the questions and answers in the books are not difficult but extremely unclear: definitions are missing, I don’t know what they mean when they wrote “expected” “forward” “future”,… what are the difference between " expected to appreciate by 5 percent ……" and " must trade at a 5 percent…" (yes, I meet this problem in real, I’m trying to decrypt it. I will post it on the weekend if I fail to understand it).
An alpha of 0.9124 is illogical. Look, an alpha of 0.05 means a 95% confidence that the calculated parameter is equal to true value, an alpha of 0.20 means a 80% confidence and so on. This is why we “commonly” set a confidence level using a related alpha (0.05, 95%). If you want, you can read further about this out of the curriculum.
I think you’re making a mountain out of a molehill, so to speak. Harrogath was trying to alleviate the concerns of the poster (the exam will be more clear).
To your question as why alpha is commonly 0.05-- many fields in research set this as their threshold, the standard. There have been psychological studies that indicate people are more comfortable with 5% chance of an error, so it’s not unfounded. A lot of it is an old tradition that dates back to the origins of the methodology. You will be hard pressed to find someone who thinks an alpha exceeding 0.1 is acceptable in common use (with the exception of some methods, like variable screening techniques, but even then, you need logic for doing so).
Harrogath is absolutely correct, though, that picking alpha should be a balance between benefits and costs (risk of making type 1 or type 2 errors, and the benefits of making a correct inference… there are many other considerations, too, but I’ll leave those for another time). I guarantee you won’t find a real statistician (or even someone with a truly good understanding of statistics) that would disagree with that statement.
Ok, so an alpha of 0.0876 (I meant 0.0876 when I wrote 0.9124)?
If you think we “commonly” set a confidence level using a related alpha (0.05, 95%), could you please explain for me why in Appendix A, Student’s T-distribution, there are values of level of significance 0.001 or 0.01 or 0.02 or 0.2 . They want to waste paper and ink, right?
If you do as you told, if someone tell you make a test, you don’t hesitate to put alpha = 5%, do you?
If I were you, I demanded him which alpha he wants to use.
It’s great if you can post “psychological studies that indicate people are more comfortable with 5% chance of an error”, I really don’t know it. People in this forum can learn a new knowledge.
And please answer my question above: why there are alpha of 0.001, 0.2,… in Appendix A.
But even you can have a “psychological studies”, this doesn’t mean you can make alpha = 5% as default value of all tests.
I don’t disagree about the existence of this balance. But it is not because of existence of the balance, you can make others guest what alpha having the best balance.