# DW test question

if DW value is 1.99 D_L 1.6 and D_u is 1.8 *CFAI talks nothing about 4-D_L and 4-D_U.* my question is if DW is less than D_L---->positive serial correlation if DW is greater than D_U—> o.O??

If DW > 4-Du, then negative. If DW > 4-Dl, then inconclusive. If DW < 4-Dl, then no serial correlation. I think.

From all the exercises i did in both Schweser and CFA material, what you suggested passme is correct, even though what Grumble listed is what Schweser recommends. This is one topic that you shouldn’t try to hard to understand, if DW is less, than positive serial correlation, if DW is greater, than negative serial correlation. So for your example, its neg. correlation.

yeah…so in this case there should be no serial correlation…fail to reject H0…imho

ya, I was referring to a similar Q in the CFAI sample. I did it the schweser way and got screwed… Thanks phrenchy.

between D_U and 4-D_U is “no serial correlation”

Okay sorry I think I am getting screwed the Schweser way too…I just assumed the CFAI sample solution was wrong. Damnit!

Think about the formula -> DW = 2(1-r), where r = the correlation of the errors with the lagged errors. If the correlation is zero, the errors are uncorrelated (no serial correlation) and the DW stat = 2. Plugging in 1 (perfect positive correlation & positive serial correlation) & -1 (perfect negative correlation & negative serial correlation) gives you DW scores of 0 and 4 respectively. Anything below dl and above 4-dl would be significant for serial correlation. There is no difference between how Schweser and CFAI present this concept, the CFA just does a piss poor job of explaining it.

@passme - also note, that while this is total bull, CFAI tells you how to test for negative serial correlation (4-dl) in footnote 49 on page 383. I would be in disbelief if they tested on a footnote to a concept in the actual CFAI texts, but would not put it past them.

What is the answer on this one? DW = 1.99 indicates that the serial correlation of the errors is very very small positive number. However DW > DH so we fail to reject null of no serial correlation. So I am not sure if we have serial correlation or not, if we do, i do know that serial correction is EXTREMELY tiny. I don’t know if DW says anything about the strength of Serial correlation.

Ok, so just to get this clarified: this is what Schweser advocates: a) 0 to D_L --> “positive autocorr” b) D_L to D_U --> “ambiguous” c) D_U thru 2 to 4-D_U --> “no autocorr” d) 4-D_U to 4-D_L --> “ambiguous” e) 4-D_L to 4 --> “negative autocorr” so in the question above --> D_L 1.6 and D_U is 1.8, and DW = 1.99 (given) This puts the data into band number “c)” amongst the 5 steps above. Hence no autocorr. Please correct me if I am wrong. Thanks.

thx guys for yall inputs. I was talking about a question in 2011 sample exam (the second one). The answer didn’t make sense to me at all. Answer: Because DW>D_u, we fail to reject the null hyp that no significant SC exist. When serial correlation is present, the SE of the estimates are artificially small. “we fail to reject” means we accept. So we accept the hypothesis of no significant SC. So there is no Significant SC. so, the answer contradicts itself. my conclusion: I spent enough time on this #!@! question and I am going to study ethics instead. Thanks for sharing.

meshed Wrote: ------------------------------------------------------- > Ok, so just to get this clarified: this is what > Schweser advocates: > > a) 0 to D_L --> “positive autocorr” > b) D_L to D_U --> “ambiguous” > c) D_U thru 2 to 4-D_U --> “no autocorr” > d) 4-D_U to 4-D_L --> “ambiguous” > e) 4-D_L to 4 --> “negative autocorr” > > > so in the question above --> D_L 1.6 and D_U is > 1.8, and DW = 1.99 (given) > > This puts the data into band number “c)” amongst > the 5 steps above. Hence no autocorr. Please > correct me if I am wrong. Thanks. hey bro, if you have access to 2011 sample, look at Q 11. I got trained by schweser on this section and thats exactly what i did and got screwed over

@passme: i did exactly the same thing, and got screwed over the same sample question…i just assumed the cfai answer was wrong…however, if there is indeed a difference between CFAI and Schweser, then I’d like to have it clarified. Can someone please help? There are a million places to make mistakes on the entire curriculum, but something as easy as this definitely shouldn’t be one of them someone pls help I tend to ignore all the DU DL rules completely as all the examples that I have seen doesn’t complicate the issue. DW of 2 means no autocorrelation (so i always assume anything between 1.8 and 2.2 means no autocorrelation) DW of 0 means +ve SC so anything close to zero for me has +ve SC DW of 4 means -ve SC so anything close to four for me has -ve SC I have used this rule in all my sample questions and i tend to get them right. I couldn’t get my head around understanding all those different classifications

Just to clarify, in no way does the answer contradict itself…you guys need to stop thinking answers are wrong lol…Failure to reject the null hypothesis implies there is no serial correlation since you’re failing to reject…remember the null hypothesis is No autocorrelation so if you’re failing to reject it then you stay with the null and hence no autocorrelation…DW of approximately 2 should be a sheer give away that there likely isn’t going to be a problem with autocorrelation by the way since a DW of 2 is typically what you hope for so you don’t even really need to waste time looking into the numbers

@MFIN - I believe he is saying that the answer was serial correlation is present, and the explanation was that we fail to reject the null. The answer contradicts the explanation.

the answer is correct. there is serial correlation. it is positive. the entire dataset is accurate along w/ the answer. i got it wrong also… this sample set is ridiculously hard.

Guys, this sample is a total garbage. If you fail to reject the null hypothesis of no positive serial correlation, then the hypothesis stands, meaning there is no positive correlation. The rest of the explanation contradicts that statements. Whoever was typing these bull was under heavy influence or smthg.