[johnpersico]

Many people on this forum could really benefit from some classes in statistical analysis. The discussions and arguments over dyno results are a good example of not understanding data and measurement fundamentals. Some basics:

1. There are no absolute values for anything, it all depends on how, when, where and what you use to measure something with. Thus, the time, weather, equipment and even person doing the measuring can and will create measurement errors and bias.

2. The only real "value" for anything is a composite of values which create a bell shaped curve, wherein we assume "True Value" can be found. The standard deviation is a measure of the range in which we can expect our "true value" to be found. The SD is a number that is based on some statistical probability and itself can be subject to error. In other words, it can be wrong.

3. The "real value" of horsepower for a bike will depend on many factors on any given day: Temperature, humidity, type of gas used, altitude, not to mention the tuning of the bike.

4. If we use the principle that economists use: "Ceteris Parebus" than we will find that some bikes are more powerful than others, but think of this as possible along a bell shaped curve and with the power bands so close between bikes, it is highly likely that there will be Gixxers more powerful than Kawis, Kawis more powerful than CBR's, CBR;s more powerful than R1s and R1s more powerful than Gixxers and so on an so forth. Since the range of "true values" for hp for these bikes will very likely overlap. My R1 will make more hp than some Kawi's and less than others. Again, I am invoking the principle of Ceteris Parebus.

5. Another factor is the error of the dyno's. Just as any survey has a range of error and we expect the true value to lie somewhere in this range, we need to ask, what is the calibration error of the instrument we are using, i.e, how accurate is it? I have not seen one dyno posting with this figure and yet it is critical as well as asking what is the range of error, since the error itself could vary from one dyno test to another even on the same dyno, not to mention other dyno's. By the way, a good question if you decide to fight a speeding ticket as I did a few years ago. I made them give me the calibration records and testing records for both the radar and the cop using the radar. My defense would have put Perry Mason to shame. I had the cop looking like he could'nt spell his name. Nevertheless, I still lost the case. Moral: Can;'t beat the system.

Hope this helps but I doubt it.

 

[r1-superstar]

Many people on this forum could really benefit from some classes in statistical analysis. The discussions and arguments over dyno results are a good example of not understanding data and measurement fundamentals. Some basics:

1. There are no absolute values for anything, it all depends on how, when, where and what you use to measure something with. Thus, the time, weather, equipment and even person doing the measuring can and will create measurement errors and bias.

2. The only real "value" for anything is a composite of values which create a bell shaped curve, wherein we assume "True Value" can be found. The standard deviation is a measure of the range in which we can expect our "true value" to be found. The SD is a number that is based on some statistical probability and itself can be subject to error. In other words, it can be wrong.

3. The "real value" of horsepower for a bike will depend on many factors on any given day: Temperature, humidity, type of gas used, altitude, not to mention the tuning of the bike.

4. If we use the principle that economists use: "Ceteris Parebus" than we will find that some bikes are more powerful than others, but think of this as possible along a bell shaped curve and with the power bands so close between bikes, it is highly likely that there will be Gixxers more powerful than Kawis, Kawis more powerful than CBR's, CBR;s more powerful than R1s and R1s more powerful than Gixxers and so on an so forth. Since the range of "true values" for hp for these bikes will very likely overlap. My R1 will make more hp than some Kawi's and less than others. Again, I am invoking the principle of Ceteris Parebus.

Hope this helps but I doubt it.

Where did this info come from? URL please, I'd like to read up on it more. Thanks........db

 

[Eyespy]

Many people on this forum could really benefit from some classes in statistical analysis. The discussions and arguments over dyno results are a good example of not understanding data and measurement fundamentals. Some basics:

1. There are no absolute values for anything, it all depends on how, when, where and what you use to measure something with. Thus, the time, weather, equipment and even person doing the measuring can and will create measurement errors and bias.

2. The only real "value" for anything is a composite of values which create a bell shaped curve, wherein we assume "True Value" can be found. The standard deviation is a measure of the range in which we can expect our "true value" to be found. The SD is a number that is based on some statistical probability and itself can be subject to error. In other words, it can be wrong.

3. The "real value" of horsepower for a bike will depend on many factors on any given day: Temperature, humidity, type of gas used, altitude, not to mention the tuning of the bike.

4. If we use the principle that economists use: "Ceteris Parebus" than we will find that some bikes are more powerful than others, but think of this as possible along a bell shaped curve and with the power bands so close between bikes, it is highly likely that there will be Gixxers more powerful than Kawis, Kawis more powerful than CBR's, CBR;s more powerful than R1s and R1s more powerful than Gixxers and so on an so forth. Since the range of "true values" for hp for these bikes will very likely overlap. My R1 will make more hp than some Kawi's and less than others. Again, I am invoking the principle of Ceteris Parebus.

Hope this helps but I doubt it.

Substitute "barometric pressure" for "altitude" for more precise and technically correct factual accuracy.

 

[CDRacingZX6R]

Many people on this forum could really benefit from some classes in statistical analysis. The discussions and arguments over dyno results are a good example of not understanding data and measurement fundamentals. Some basics:

1. There are no absolute values for anything, it all depends on how, when, where and what you use to measure something with. Thus, the time, weather, equipment and even person doing the measuring can and will create measurement errors and bias.

2. The only real "value" for anything is a composite of values which create a bell shaped curve, wherein we assume "True Value" can be found. The standard deviation is a measure of the range in which we can expect our "true value" to be found. The SD is a number that is based on some statistical probability and itself can be subject to error. In other words, it can be wrong.

3. The "real value" of horsepower for a bike will depend on many factors on any given day: Temperature, humidity, type of gas used, altitude, not to mention the tuning of the bike.

4. If we use the principle that economists use: "Ceteris Parebus" than we will find that some bikes are more powerful than others, but think of this as possible along a bell shaped curve and with the power bands so close between bikes, it is highly likely that there will be Gixxers more powerful than Kawis, Kawis more powerful than CBR's, CBR;s more powerful than R1s and R1s more powerful than Gixxers and so on an so forth. Since the range of "true values" for hp for these bikes will very likely overlap. My R1 will make more hp than some Kawi's and less than others. Again, I am invoking the principle of Ceteris Parebus.

Hope this helps but I doubt it.

Dyno’s were never claimed to be completely accurate. They are supposed to give people a general idea as to what power the bike is producing. Most magazines take a series of dyno runs, at varying times of day and temperatures, etc. Then average the dyno runs together to produce (In the magazines opinions) the closest and most accurate dyno test they could run.

Lets face it, if a hundred magazines all around the world dyno the new R1 this year. And the average dyno output is 151 or so. Then the reality of the matter is the bike is making pretty close to that exactly, regardless of humidity, temperature, location, altitude. The game of statistics is a game of averages.

 

[johnpersico]

(Lets face it, if a hundred magazines all around the world dyno the new R1 this year. And the average dyno output is 151 or so. Then the reality of the matter is the bike is making pretty close to that exactly, regardless of humidity, temperature, location, altitude. The game of statistics is a game of averages.)

Not true my friend: All depends on the SD of the values. The average is not the same as the "true value." What good is an average if it can be anywhere between 100 and 200 hp. As far as sampling and statistical reliability goes, this is what your comment gets to here, I would have to have a determination as to how many samples are needed to be taken to produce a range of reliable values within say 3 standard deviation, which is a measure that assumes a 99.7 percent reliability around our assertion that the "true value" lies somewhere in the given range of values. Your comment above is not realistic, since there will not be 100 magazines reporting it. There might be 100 dynos if everyone on this forum gets their new r1 dynoed, but the statistical issues will still be valid and thus we would need to see what our range and SD is from all these dyno's and compare them to the same tests for the other new liter bikes.

 

[johnpersico]

Re: Re: Real Dyno Figures don't exist

Where did this info come from? URL please, I'd like to read up on it more. Thanks........db

Try these URLs for a start: http://trochim.human.cornell.edu/kb/measerr.htm

http://trochim.human.cornell.edu/kb/truescor.htm


I would also suggest looking under sampling bias and measurement error on Google and you should find some good information.

 

[tiller2nv]

Look at it like this, if two bikes are the same dyno at the same time then you will have a HP winner.

 

[johnpersico]

Look at it like this, if two bikes are the same dyno at the same time then you will have a HP winner.

True, but lets say that one is a ZX10 and the other an R1, this simple test would not prove that:

1. All ZX10s are more powerful than all R1s
2. That another dyno might in fact reverse the findings
3. That the hp measured is in fact the average hp of the class of bikes represented by the tests.
4. That the measurement given, lets say 159 rwhp is at all accurate for the bikes measured.

So, from a point of view of hypothesis testing, you have a weak hypothesis not suitable for generalizing to any other bikes or even to these bikes on any other dyno's.

 

[tonykart8]

This is very interesting theory...So why do race teams and mod shops use dynos? A dyno is a tool to create comparative values between two resultant inputs.. ie..R1 and ZX10rr or stock vs mod... That is real...

 

[johnpersico]

This is very interesting theory...So why do race teams and mod shops use dynos? A dyno is a tool to create comparative values between two resultant inputs.. ie..R1 and ZX10rr or stock vs mod... That is real...

Glad you asked that question: See many businesses use metrics and do not have a good understanding of measurement error, ergo, the metrics are not as useful as they could be and in some cases may even be less useful than no measurement. However, if a person understands the basics of measurement theory, they can make more reliable and valid determinations as to the implications of said theory. So if I understand that my dyno has a 3 percent measurement error and I am tuning a new setup for a race bike, then the fact that the dyno reads 161 with the new setup but read 159 with the old, will be meaningless, given the potential measurement error. However, if I take enough tests, and I find a trend (generally 12 measurements in a row) or I find a suitable sample size that allows a high probablilty that there is now a significant difference in power with the new setup, then my efforts have not been in vain. For instance, if you buy a new exhaust system and it makes 3 more hp on a dyno then your stock one, can you consider that a significant difference (ergo, the result of the new system) or could it be jus measurement error? On the other hand, if the difference is 15 or 20 hp more, statistics can tell us the probabilty that the increase is a "real" increase versus the possibilty that it is just measurement error. A very useful thing to know if you spend mucho money on parts and bike tuning. Hope this makes sense.

 

[Mackz]

Hmmm, interesting thread ...

 

[Big Daddy]

Intriguing thread indeed!! Aren't dyno tests/graphs used as a referance just as suspension settings are suggested for a said bike and adjusted/tweeked to suit??

BD

 

[afabbrin]

I'm having flashbacks to statistics - help!! :scared

j/k - this is a good thread.

 

[CDRacingZX6R]

(Lets face it, if a hundred magazines all around the world dyno the new R1 this year. And the average dyno output is 151 or so. Then the reality of the matter is the bike is making pretty close to that exactly, regardless of humidity, temperature, location, altitude. The game of statistics is a game of averages.)

Not true my friend: All depends on the SD of the values. The average is not the same as the "true value." What good is an average if it can be anywhere between 100 and 200 hp. As far as sampling and statistical reliability goes, this is what your comment gets to here, I would have to have a determination as to how many samples are needed to be taken to produce a range of reliable values within say 3 standard deviation, which is a measure that assumes a 99.7 percent reliability around our assertion that the "true value" lies somewhere in the given range of values. Your comment above is not realistic, since there will not be 100 magazines reporting it. There might be 100 dynos if everyone on this forum gets their new r1 dynoed, but the statistical issues will still be valid and thus we would need to see what our range and SD is from all these dyno's and compare them to the same tests for the other new liter bikes.

Exactly how many magazines will dyno it?

 

[rydfree]

Look at it like this, if two bikes are the same dyno at the same time then you will have a HP winner.

That's one wide ass Dyno

 

[Andy29]

So the point you're making is no one will ever know who of the 4 1000s has the most power because they're all too close?

 

[johnpersico]

So the point you're making is no one will ever know who of the 4 1000s has the most power because they're all too close?

No not exactly, you are close though. What I am saying is "If they are really that close", then the probabilty is very high that given each bike will have some normal variation, for instance, if you took 100 ZX10's, they might have a spread of lets say: 148 rwhp to 152 rwhp, if we dynoed enough of them we could say that the probability exists that the true value of the HP for the Kawi lies somewhere between 148 and 152. Now lets say the Yamaha, we do the same and the spread might be 150 to 154, we can say then that the true value lies somewhere between 150 and 154. We can thus conclude:

1. The average of the Yamaha is 152 versus the average for the Kawi which will be 150.
2. On average the Yamaha makes more hp than the Kawi.
3. However, there will be a heck of a lot of Kawi's that make more hp than Yamahas. Since, 25 percent of the ZX10s will make between 151 and 152 hp and approximately 25 percent of the R1s will only make 150-151 hp.

I hope this helps some. But you can see that my point is if the bikes are indeed so close it will take many tests before we can find reliable distributions for the hp ranges of these bikes. Only then can we make the above determinations. Based on so few dyno runs as we now have with the new bikes, they are only a start and each individual dyno run is only ONE DATA point that we need to create a reliable and valid distribution or pattern.

I will say this though, I am buying an R1 so I have a built in bias.

 

[GXRKLR]

This has been brought up but not in so many words. In the end you are still no closer than you were before. With this method your taking numbers and turning them into a generalized statement. If we wanted that we would have called the Phsycic Network. "I see the letter A, the king of HP this year has the letter A in it's name." So, it may not be that vague but you get the idea.

The only advantage to this is because it is so general you can't really argue it, and where is the fun in that.

 

[Andy29]

This has been brought up but not in so many words. In the end you are still no closer than you were before. With this method your taking numbers and turning them into a generalized statement. If we wanted that we would have called the Phsycic Network. "I see the letter A, the king of HP this year has the letter A in it's name." So, it may not be that vague but you get the idea.

The only advantage to this is because it is so general you can't really argue it, and where is the fun in that.

Some of us aren't allowed to argue for fear of banishment.:cryin

 

[GXRKLR]

Some of us aren't allowed to argue for fear of banishment.:cryin

That depends on what your arguing about and how you go about it.

But I guess you having a gixxer puts you on thin ice anyway.:lol