|
|
10-29-2007, 09:36 PM | #45 | |
Major General
374
Rep 8,033
Posts |
Quote:
It is possible that Nissan has a superb driver, state of the art transmission, AWD system, and suspension all in one car, and therefore manages to escape the trend. If the GT2 can do it, and since the GTR can be expected to be more advanced than the GT2, it is possible that it raises the bar slightly further. That doesn't invalidate the outcome of the analysis--that there is a strong linear relationship between weight/power and lap time on the Ring. Last edited by lucid; 10-29-2007 at 09:57 PM.. |
|
Appreciate
0
|
10-29-2007, 10:20 PM | #46 | |
Lieutenant General
609
Rep 10,407
Posts |
Keep digging
Quote:
Whether you like it or not the regression is what it is. lucid did not regress against any straight line performance metrics (0-60, 0-100, 1/4 mi, etc.). The regression was for the Nurburgring lap time, period, and the results speak for themselves. Lap time is largely explained by power to weight ratio. Other factors are present and visible in the data and many, such as the effect of race tires are cleary visible and explained by the data in terms of distance below the regression line. Of course there are some hidden correlations in the data as cars with a better power to weight ratios will typically by design (almost by definition...) have better suspension set ups, wider tires, etc. Nonetheless, much of this variation can be seen as the spread of data at a given point along a given narrow columnar section of the plot at a given power to weight point. As I mentioned it would be fantastic to have a similar analysis for other dominant effects such as tires, transmission, suspension design, center of gravity, redline, etc. but my strong suspicion is still that power to weight will be the largest effect by far. Of course the slope of the line and perhaps the R^2 value would change track by track with tighter slower speed courses favoring light weight cars with very dialed in handling. However, when because the Ring is simply largely a test of a cars ability to accelerate you are left with the excellent R^2 value. Thinking that all of the distance the GT-R point holds as an outlier being explained purely by suspension goes against the data as well against common sense. Could Nissan have entirely reinvented the track/race car suspension system? Is this likely? Occam's Razor.... . and hence the appropriate nature of the previous comment about "UFO like handling". So this is possible, just extremely unlikely. Perhaps you can quanitify your speculation that the entire time could be due solely to the suspension? I'm happy to examine your analysis that shows this is a possibility - anything other than an analysis is pure speculation. On errors: I would be happy to quantify the nature of the GT-R data point in terms of probabilites if lucid shares the xls file with me. However, all it is going to do is give us a probability. That probability may be 40% or it may be 99%, who knows, further once we have it it still won't really settle this debate. On the GT2 time: Sure it is an outlier as well and that is explained perfectly well by realizing the car had Cup+ tires good for about 15 seconds on this course. Point, set, match on the GT2. There is no insulting going on here. I have said it once, probably twice and I will say it again. This represents a remarkable achievement obtained by a remarkable car. Nonetheless I will stick to my hypothesis as previously stated and will place money on that if you are so confident I am BS-ing. It is funny how very healthy skepticism and a bit of math makes some folks so quick to call BS. All such car debates could use a very strong dose of statistics and physics. |
|
Appreciate
0
|
10-30-2007, 01:02 AM | #47 |
Major General
1094
Rep 8,013
Posts
Drives: i4M50
Join Date: Oct 2007
Location: No where fast
|
From the brief experience I have had on the ring, suspension/ride plays a very big part in how a car performs, too stiff a suspension and the car bounces from bump to bump. It's hard to keep grip constaint.
If the Carrera GT had a slightly softer suspension it would have been way better on the ring, but with the same token would have been worse on the normally smooth race tracks which it was intended for. The ring is only really useful for tuned street car, not semi/race cars. Last edited by footie; 10-30-2007 at 11:31 AM.. |
Appreciate
0
|
10-30-2007, 08:52 AM | #48 |
Captain
17
Rep 658
Posts |
here's the original link on edmunds, if you scroll down to the heading for "Red GTR" you'll see it.
http://blogs.edmunds.com/Straightlin...ory/cat.Nissan |
Appreciate
0
|
10-30-2007, 01:04 PM | #49 | |
Major General
374
Rep 8,033
Posts |
Quote:
|
|
Appreciate
0
|
10-30-2007, 09:16 PM | #50 |
Private First Class
6
Rep 123
Posts |
That's exactly the problem. What's great about you (and swamp as well -let's not leave him out cause he'll be pi$$ed) is that you take issue with what I said even when you don't see my point. That's raw genious at work!
To my point. Nobody disputes that horsepower is an important determinant of laptime. Certainly not me and certainly not anyone I know. To see this clearly consider a car with an engine that produces zero horsepower. The car will not move at all and its laptime will be infinite. I think we also agree that there are other variables that contribute strongly to laptimes. But the regression line you fit to the "normalized" power data doesn't contain any of these other variables. That's OK you say, as long as the regression error bands from all OTHER cars I have already charted contain or "bound" any new cars, I am not going to question their laptimes. But when one new point of data arrives that is slightly outside the current error bands your world is shattered. Something is amiss you say, my single parameter regression model doesn't fit as well so something is wrong with this picture and I am going to question it (that's what swamp says at least). This is exactly my point. YOU CANNOT INFER ANYTHING from this model about your new data point. Nothing, nada. Neither can swamp (even if he thinks otherwise ) To do so you require additional information which YOU DO NOT HAVE! Oh, by the way, I am completely relaxed ... |
Appreciate
0
|
10-30-2007, 09:23 PM | #51 | |
Captain
13
Rep 689
Posts |
Quote:
Suspension and tires (among other things) matter a lot more than HP. |
|
Appreciate
0
|
10-30-2007, 09:58 PM | #52 |
Major General
374
Rep 8,033
Posts |
What are you talking about? What world? How is what shattered? All I've said so far is that the analysis identifies it as an outlier--the most extreme among the 75 cars in the analysis which include all sorts of cars with excellent suspension/handling characteristics. Re-read the part of my post (#45) that you summarized with the "blah blah" and you'll see my take on that. What would your take on the GTR time be if the car had 300hp? You'd still say it has excellent handling and it's possible? Of course not. So, we are talking about the dimishing chances of an event occuring as you move away from the regression line--provided regression has been established, and in this case, it has been. That's basic stats. How can one dispute that? You still are missing the point of running a regression analysis. Whatever I guess...
|
Appreciate
0
|
10-30-2007, 10:20 PM | #53 | |
Lieutenant General
609
Rep 10,407
Posts |
Same back at ya
Quote:
I am not questioning the lap time, you saw it just like I did. That is a big part of my point, SOMETHING caused the point to be an outlier (unless we are back on the magic or the UFO hanlding theory). Hmmm... what could those somethings be. Rather than throw my hands in the air, hopelessly claiming we can not know anything as you have done, I'll offer a list: 1. Power 2. Weight 3. Handling prowess (or handling black magic...) 4. Tires 5. Transmission performance 6. The driver 7. Extreme environmental conditions You rule out 1 and 2 by saying Nissan could not, would not, does not have a sleeper (I refuse to rule it out, hence my ancient #1-#4). We can probably rule out 7 as well. What is left, handling (including AWD, traction control effects, etc.), tires, transmission and driver. I am convinced the DCT transmission did have an effect and perhaps sizeable, maybe as much as 10 seconds. I am looking for data to show how large this effect may be but am empty handed at this point. My 10 second figure was based on counting power on upshifts in a lap and doing some math but the estimate is very rough (better than nothing - hint, hint). Where does the remaining 10 seconds come from? I am not convinced that Nissan has reinvented sports can handling so I will call that one as possible but unlikely. How amny other high tech AWD cars with advanced traction control are represented. I have not counted but probably a couple at least. I guess this is where you and I disagree. You really have not said so explicitly, yet and my challenge to you to demonstrate some evidence that suspension alone could be responsible remains. I see you have come up 100% empty handed on this. Enigma did say that is his opinion, but again no evidence. Without evidence I call it a hunch or worse, speculation. Certainly the driver as well as tires will both have reasonably large effects as well. Since some of the other lap times include very talented drivers I'm also suggesting the driver is not repsonsible for much of the remaining nature of the point as an outlier. What is left, TIRES. I think I will still stick with my previous #1-#4 with the slight revision posted along the way. I'll also go out on a limb and speculate that if it is not #1-#4 or some combination thereof it is then a combination of DCT, some massive innovation in handling and the driver (a total ace/god). You always should consider Occam's Razor as well. I don't suppose you though about that? Lastly you keep saying how important the other or unknown factors can be. Have another quick look as the regression plot. The power to weight factor varies by about a factor or 4, that is enormous. The car design and most ways of describing the cars varies enormously. The lap times correspondingly vary by 100 seconds (close to 2 minutes). Do you think your "other factors" will correlate across such a broad range as this and with as good of an R^2 value? Power to weight is #1, period. |
|
Appreciate
0
|
10-30-2007, 10:35 PM | #54 | ||
Lieutenant General
609
Rep 10,407
Posts |
-1, +1
Quote:
Quote:
On the tires point I can't quite tell if you and coaster agree or disagree?? |
||
Appreciate
0
|
10-31-2007, 01:08 AM | #55 |
Second Lieutenant
8
Rep 278
Posts |
I thought I'd add one more thing to help the less fortunate understand the regression more clearly:
The regression is using both hp AND weight. (Not just hp) Two variables. Count em. Not one. But twoooo... Seeing as how these two variables are the first and second most important variables, it is perfectly natural that the R square is so high and the regression is so powerful. That being said, lets all move on If lucid and swamp would allow, I would summarize the statistics position as this: GTR lap time exceeds expectations by a large amount making it extremely unlikely (statistically). However, given that the DCT will reduce its time, and Superior tires will reduce its time: Nobody is denying that the time is possible. Just saying that its lap time is far superior to the competition given its hp and weight. But since it is employing a revolutionary transmission and may have used superior tires, its time is more than possible. The regression is simply underscoring the fact that something special is going on here cuz the kg/hp is nothing special. (AKA it has DCT, and tires maybe) Now, lets talk about the fact that the GTR can transfer power from side to side in addition to the usual front to back. Is this true? Wouldn't this create huge acceleration gains when cornering? |
Appreciate
0
|
10-31-2007, 09:28 AM | #56 | |
Private First Class
6
Rep 123
Posts |
Quote:
And yes we get it that it's the #1 variable affecting laptimes. What you "fine" gentlemen completely miss and continue to insist on is that this single-parameter regression ERROR BANDS are somehow useful in questioning any new data points that fall barely outside the bands. Since your statistics background as it relates to inference might be a bit weak please allow me to suggest the following book: http://www.amazon.com/How-Lie-Statis.../dp/0393310728 If I got a dollar for every BS statement I have heard over the years that was backed up by improper use of statistics I would be ... let's say better off than I am today. Enjoy your day and let's move on to a new topic until more info has surfaced on the GTR. Word of advice to swamp: "speculation" is like a drug. If you are realy good at it try the stock market. |
|
Appreciate
0
|
10-31-2007, 09:47 AM | #57 | |
Major General
374
Rep 8,033
Posts |
Quote:
It's amazing that you cannot calm down and actually try to understand what people have been saying about outliers--that they are merely outliers and that simply means that there is less of a chance of them occuring than an event that falls within the confidence interval. Instead of engaging in trash talk, you should prove to us why that is not true, which is basically what you are saying. |
|
Appreciate
0
|
10-31-2007, 09:53 AM | #58 |
Lieutenant
35
Rep 563
Posts
Drives: 2007 E92 M3
Join Date: Mar 2007
Location: London
|
I really don't think we need to get our respective knickers in a twist here.
IIRC the original regression came back with a P-value < 0.001 and I expect the subsequent iterations are similar. The significance of the correlation is therefore established - we don't really need to keep arguing about the simple but proven method (and one-upmanship about statistics of all things is a bit pathetic anyway). Regressive methods by definition cannot establish a cause and effect linkage. We have also established that certain cars do not strictly conform and can make educated guesses that the reasons for this are: 1) tyre compound 2) philosophy of design 3) quality of engineering The real question is: "Does the new M3 have enough of the special ingredients that made the E46 CSL perform so well at the 'Ring when compared to other M-cars?" |
Appreciate
0
|
10-31-2007, 02:47 PM | #60 |
Lieutenant General
609
Rep 10,407
Posts |
One small correction
Such methods may not establish cause and effect, but to do show it's existence, sematics perhaps. Let's imagine that for some strange reason F=ma was not known nor understood and it was not obvious that better power to weight = faster on a track. This regression would show such cause and effect. Sure a formula is a better indication of cause and effect but statistics show the same thing as the formula.
|
Appreciate
0
|
10-31-2007, 03:18 PM | #61 |
Major General
374
Rep 8,033
Posts |
Regression is usually a first step in an exploratory analysis that is intended to contribute toward future work that is focused on identifying cause and effect relationships. The outcome of the regression, once the model meets certain quantitative criteria, is explored qualitatively, and cause-effect relationships can be speculated as hypotheses. Additional and more specific experiments/analysis are almost always needed to test the validity of the resulting hypotheses. All we’ve done here is the first part.
|
Appreciate
0
|
10-31-2007, 03:19 PM | #62 | |
Lieutenant
35
Rep 563
Posts
Drives: 2007 E92 M3
Join Date: Mar 2007
Location: London
|
Quote:
But if you take a different kind of retrospective study, such as the one published today that proved a significant correlation between obesity and malignancy, proof of the correlation does not determine cause and effect. Does being fat increase the risk of cancer? Or do people with cancer get fat? I did say that arguing about stats was sad, so I'll leave it at that... |
|
Appreciate
0
|
10-31-2007, 11:16 PM | #63 | |
Second Lieutenant
8
Rep 278
Posts |
Quote:
Seeing as how I actually PERFORMED this regression on Excel (like lucid), I am quite aware that they were normalized. Back on track The only way the GTR is not bounded by our analysis is if it has something superior to every single car in this analysis. For instance: 1)revolutionary suspension (with UFO handling characteristics) 2)DCT If you think the GTR suspension is far superior to any other car included in this analysis, then you could argue that it is somewhat meaningless. You keep talking about the importance of suspension, so Do you really think that is the case? That no other car included in this analysis had a suspension as good as the GTR?? Not one? The GTR's weight and hp are both bounded by this study. AKA the regression is using cars with just as much or little hp, and is using cars with just as much or little weight. If the GTR was the lighter than every single car in the bunch, then the regression would not be able to predict the GTR's expected time. But since it does not have extraordinary hp or weight, then this regression absolutely applies. We therefore KNOW that the regression applies regarding kg and hp. We are assuming that it does not have UFO handling therefore the regression applies. And we are also assuming that Mizuno is not an alien with driving ability which exceeds any other test driver of any of the cars used in the analysis. For the last time, that is why the conclusion of this regression is that the GTR significantly exceeded its expected time given its kg/hp and that it does not have UFO handling(vs. the cars in the study) The distance from the least squares line (21sec) must therefore be explained by something other than hp, kg, suspension. Which is why swamp made predictions as to what that something is. And it is looking like it is a product of DCT, tires, awd? We dont know yet. I don't know what assertion is bothering you so much? The time savings of DCT and tires will perfectly explain the GTR time. No one is trying to lie about these statistics. When someone eventually quantifies the advantage of DCT and the tires being used, my guess is the 21 sec error will be perfectly explainable. If it did not use special tires, then we can either infer that DCT alone saves 21 sec or that the GTR has something else never before seen. If you would like predict "amazing suspension" then by all means. Swamp predicted 4 possibilities. I predict DCT and tires. The regression is solid. It has almost scarily predicted the advantage of DCT and of tires. You really think we can't apply the regression to the GTR and conclude that it has something extra worth 21 sec of advantage? The chances of that statement being disproven by future findings is slim to none. What is hilarious is that if DCT and tires are, in fact, being used, and they are found to result in time savings of OVER 21 sec somehow, then the regression would suggest that the GTR actually has poorer than avg handling characteristics(vs the cars in the test) or that mizuno is a bad driver, or that the course was wet, yada yada. When more is understood about the GTR lap time, I think you will see that the regression was actually dead on.(predicting that something funny is going on AKA DCT, tires, etc)(Like SWAMP predicted) On that note, I don't think anyone wants to talk about the regression anymore because of how critical you have been of it. |
|
Appreciate
0
|
11-06-2007, 12:17 AM | #64 |
Lieutenant General
609
Rep 10,407
Posts |
Horses (beating)
Guys, not trying to beat a dead horse but I found this thread very interesting. After digging around a lot tonight online to try to get answers I only find arguments continuing. Some claim the Nissan Chief Engineer said the car was on certain tires, refuting statements claim the opposite. Heck ther is even debates about the actual quotes, which sites have the quotes and is which cases it is only forums/blogs. There is still some debate about the length of the particular run (maybe that one is actually settled...). For sure the car is still a pre-production vehicle. There are also stories about past Nissan Skyline N'ring lap time exaggeration ... and yet again stories refuting such claims about the Skylines. Ugh. Not sure this will be "answered" until magazines get production cars. Still then maybe it won't be answered because Horst surely is not as fast and talented as the head Nissan test driver.
|
Appreciate
0
|
11-06-2007, 03:26 AM | #65 |
Major General
1094
Rep 8,013
Posts
Drives: i4M50
Join Date: Oct 2007
Location: No where fast
|
In this month's CAR magazine in is stating that the GT-R has done the ring in 7:38 so does that mean it's official.
|
Appreciate
0
|
11-06-2007, 07:18 AM | #66 |
Expert Road Racer
59
Rep 1,329
Posts |
Here's the best post on the GT-R I've found from a guy who claims to have inside knowledge:
Well, this has been a very entertaining thread - one that I typically would not get entangled in, but let me put my 2 cents in and then shut myself up. As an mechanical engineer by education and an engineering manager by professional, I am constantly asking my engineers to show me the data to support their claims or conclusions. It seems that Nissan with their "viral marketing" campaign with the GT-R, heard loud and clear with threads like this very one, to show us some data to support the Ring laptimes. So what do they do? They open their official GT-R unveiling without any spoken words, but with a full lap in-car video which closes with the time of 7:38:54 - let the data talk. As a car enthusiasts, I was quite impressed with such a display, but the ending did surprise me. Why? Well as an occasional contributing editor to some auto related sites, I was privy to Nissan's GT-R presentation a few days earlier in which we had the opportunity to talk to the engineering team and some of the Nissan's development partners for the GT-R. During Mizuno-san's opening presentation, Ring times of 7:35 were mentioned quite a few times. So I was was expecting a time of 7:35 to show up at the end of the video. I can respect the philosophy of conservative specifications with overachieving performance - the speak softly but carry a big stick mentality. No one wants to go through the type of embarassment that Ford did with the over rated horsepower ratings of the Mustang a few years back. More data - At the presentation, Nissan made note of all the data logging that was done with the GT-R. They had on display the datalog for 2 complete laps at the Ring. These particular datalogs showed 7 parameters: speed, X direction G-force, Y-direction G-force, throttle position and brake fluid pressure. One of the datalogged laps was the 7:38:54 lap on September 24 and the other happened to be one for a 7:37:50 on September 20. The interesting thing about the 7:37:50 lap is that there is 2 instances where they were slowed by other test cars. The data of both the 7:38:54 and the 7:37:50 laps are superimposed on each other, and one can clearly see a drop in speed when the encounters were made with the other cars. On this faster lap, there were no wet sections of track and the data does show that they carried more speed through those particular sections. So why not show the faster full dry lap or even the 7:35 lap that was mentioned in the presentation? Did Nissan want to show a video of one clear lap with no other cars on track? Are they playing it conservative, as they know that future independent testing will be done? My pure speculation is that they wanted to show something just fast enough to beat their intended competition and yet give themselves some breathing room for the independent testers. Furthermore, as was made very clear throughout the presentation, Ring times were only one part of the balance equation for the GT-R and they easily surpassed their initial early targets for the car. Ok so we have the times, but what was there any special set up for the car and will others be able to validate Nissan's times? Well we did ask that in a long video taped interview of Mizuno-san (I am the fat old geek asking most of the questions). I specifically asked if there were any special tires, brake pads and engine tuning for the Ring laps. He clearly states that the car is stock. Yes, I did see the other comments about cut slicks, but I speculate that that has to do more with a misunderstanding or communication issue. As I now live and work in China, I am now so familiar with the issues of language miscommunication and encountered it many times while interviewing quite a few of the Nissan executives, including Mizuno-san, during the TMS event. In talking to other engineers and also some of the development partners, I again speculate that the conservative approach is the path being taken. Are horsepower ratings, shift times, lateral G's, etc, conservative as well? Well let's see what the independent testers come up with or better yet, give me one long term loaner to test for a year. Does this make things any clearer? Maybe, maybe not? As with any data, both sides will spin the data to whatever supports their views or stands. So just wait for the next edition of Nissan Sport Magazine and look and review the data and other information and draw you own conclusions. Glenn PS: I vow not to return to this thread.... Found here: http://www.nagtroc.com/forums/index....=100&start=100 |
Appreciate
0
|
Post Reply |
Bookmarks |
|
|