Radius

[jbrock27]

ACY, it is not that I don't believe you, I do.  What I don't believe is that it is problem that cannot be explained or better yet, corrected.   You have yet to take the time to answer my questions I put forward to you, but are taking the time for this back at forth.  I don't know why.  Suit yourself.  You did say you put speakers in these Athearn SD45s-you don't remember what you saw when you took the shells off??  

If you would like to be so thin skinned as to take offense, that is your choice, just as it is to assume what you are saying to be an absolute, when others have stated they have had a different experience.  Why would what YOU have to say about the matter, be treated as more meaningful than when someone else says they have had the opposite experience?!?

You're right, I don't know your LIONEL boy's background.  But what I do know, is I don't have any problems running the  3 six axle diesels we have on 18"R curves.

[ACY]

ACY, it is not that I don't believe you, I do.  What I don't believe is that it is problem that cannot be explained or better yet, corrected.   You have yet to take the time to answer my questions I put forward to you, but are taking the time for this back at forth.  I don't know why.  Suit yourself.  You did say you put speakers in these Athearn SD45s-you don't remember what you saw when you took the shells off??  

If you would like to be so thin skinned as to take offense, that is your choice, just as it is to assume what you are saying to be an absolute, when others have stated they have had a different experience.  Why would what YOU have to say about the matter, be treated as more meaningful than when someone else says they have had the opposite experience?!?

You're right, I don't know your LIONEL boy's background.  But what I do know, is I don't have any problems running the  3 six axle diesels we have on 18"R curves.

1. I installed sound in the one SD45 in 2006 or 2007. I don't recall exactly. I currently have most of my locomtives and rolling stock in storage, so checking won't be easy.
2. What I have to say or another person on the forum has to say holds equal weight, perhaps their opinions should be more highly regarded. However, what I posted was not an opinion; it was the results of the tests that were performed on 10 Bachmann consolidations.
3. The person who was with Lionel is in his seventies and now retired as far as I know. Also I have not been discussing 6 axle diesels specifically. I only cited the results of the tests performed on the consolidations as an example.

You are more than welcome to perform your own tests on a larger sample size and see if you get contrasting data, or data to support your opinion in other words. I used statistical calculations on the 10 consolidations I tested to find that at a bare minimum at least 5% or up to 40% can have issues running on 18" radius from that given production run. I utilized a 2 tail t-test on my data to obtain these results.
I would share the full results if I could but I spilled water on my keyboard this afternoon so my computer is currently not working so I have to send it in for repair most likely.

[Jerrys HO]

I set up an oval with 18r. I ran 3 SD40-2's, 3 SD45's, 1 SD70 for 10 laps each.
None derailed or had any problems.
Does this mean I can assume that 100% of the SD's I purchase will run on 18r or just 70%?

I still back JBrock unless I happen to stick my wet finger in a light socket .

It all depends on track work and how well the engine and rolling stock are  in gauge and adjusted. I may believe 1% of loco's and rolling stock are a little out of tolerance, 30% and there would be a lot more problems than this thread has noted.

[ACY]

I set up an oval with 18r. I ran 3 SD40-2's, 3 SD45's, 1 SD70 for 10 laps each.
None derailed or had any problems.
Does this mean I can assume that 100% of the SD's I purchase will run on 18r or just 70%?

I still back JBrock unless I happen to stick my wet finger in a light socket .

It all depends on track work and how well the engine and rolling stock are  in gauge and adjusted. I may believe 1% of loco's and rolling stock are a little out of tolerance, 30% and there would be a lot more problems than this thread has noted.

I conducted the tests without rolling stock. If you are able to test more than 10 identical locomotives (from the same production run), then your results will have statistical significance. I was barely able to establish any statistical significance with 10 identical locomotives. That is why with my results I was only able to establish with 99% confidence that 5% would perform in this manner even though 3 out of 10 did in my tests. The larger the sample size, the greater the significance of the results and this directly impacts the confidence interval and other related statistical data.

The tests were performed on 3 different 36" circles of Bachmann EZ track each assembeled multiple times. The track was brand new from three different sets I had purchased recently. An NMRA standards gauge was used to check to make sure all the wheels were in guage.

Although it would be easy to say that a certain locomotive had this or that problem that could be rectified; you will have to trust my word that no issues or differences could be found. I would not trust my ability, which is why I relied on others assistance to ensure nothing was missed or overlooked.

Also consider that many people and clubs have layouts with 22" minimum radius or larger so even though a certain percentage may have trouble with 18" radius, it is possible many people never knew since they never attempted to run their locomotive on anything less than 22" radius. Or even possibly 20" radius if they use flex track.

It would be nice to have the opportunity to perform a larger test with a sample size, but it would be very time consuming and difficult.

[Irbricksceo]

Alright everyone, we're straying away from the question. to the OP: I would still recommend the 2-8-0, its a popular locomotive, easy to find, and nice looking. That is for steam. Diesel-wise I'm partial to the Geeps but I don't model Diesel.

Regarding odds, to weigh in on this 30/70 thing, if we expand the sample size and look at the possible range, we learn that it is impossible to predict. for example, lets assume that the inability to run on 18 is a defect caused by some miniscule and hard to notice wheel eror in manufacturing. Lets also assume a run of 5,000 for ease. ACY said he tested 10 and 3 failed right? well that is 30% of the tested but if we look at the sample, it could be that 4993 were correct, or it could be that 7 were and 4993 weren't the fact is, we have no way of knowing FOR SURE. Furthermore, even if we take it as a proper representation, the whole "i have 7 so the next three must be bad" doesn't work, not only because of the aforementioned Gamblers fallacy, but because we are testing from drawing one out the the pool of locomotives. even takign the (ridiculously high) failure rate of 30%, that means that there were 3500 good ones and 1500 bad ones. Youd odds are ACTUALLY based on the 3493 remaining good ad 1500 remaining defective (since we do  not know where they went) meaning you STILL have a higher chance of pulling a good one. but since there could be 3-4993 bad ones per run, we cant say.

Fun fact, my grandfather used to work in the Lionel repair shops in NJ. Oh the stories he's told.

[Jerrys HO]

I just wonder what percentage was built on Monday and what percentage was built on Friday .

[jbrock27]

ACY, I don't need to run any tests or produce any statistical data to support what I know and experience-no problems running 2 SD45s, 1 SD9 and 1 SDP40 (all Athearn) on 18"R curves, 100% success.  If you would like to send me some locos to try out for further testing/fixing I would be glad to check them out and run some tests. Also, if you ever take the time to bother taking those SD45s you have outta storage, get back to me.

It is unfortunate that you are still missing the boat here; you have had your problems running the same locos that others here have reported having no problems running, on 18"R curves.  That should tell you something, but somehow, it is not, nor does it look like anything said is going to change that for you.

**Please note, I am not making any statements here about Consolidations, nor have I in any prior posts.

[jbrock27]

I just wonder what percentage was built on Monday and what percentage was built on Friday

Yea, I heard about that kinda thing .   In the spirt of Halloween, I wonder if any of those bad locos, had serial #s that ended in : 666

[Doneldon]

Jim-

As part of my main vocational training as a clinical psychologist I had to take a load of statistics and experimental design classes -- enough that I have the equivalent of a master's degree in statistics. That's how I'm able to look at ACY's data and discuss the gambler's fallacy.


ACY-

I found your statistics on the go/won't go question about Bachmann Consolidations on 18-inch radius curves to be interesting and worth playing around with. However, your more recent post reveals that you are taking what are essentially anecdotal findings and attempting to predict from them. Your data do not support that. Further, your recent use of that data demonstrates a lack of understanding about how to use statistics and how to properly design an experiment. Let me try to help.

You need to be careful of interpreting results beyond what your data show and, especially, reporting post hoc confidence limits. Yes, I know that rule is routinely violated in the academic literature of many disciplines, including psychology at times, but the most rigorous journals still don't allow it. That's part of why certain journals are so prestigious; they have high standards and they stick to them.

Your sample is so small and uncontrolled as to be intriguing but not meaningful. The size of your sample, as well as its lack of stratification, renders it worthless for predicting anything about the entire population of Bachmann 2-8-0s. There are just too many parameters for which you haven't controlled. For example, do you know if all of your Connies were built in the same factory by the same workers? Do you know how long each sat in storage in a Chinese warehouse, U.S. warehouse, distribution center or retailer's back room? Do you know if any of your locos were subjected to unusual treatment during shipping? Do you know if any construction changes were made or whether each of the locos was built to precisely the same plans as part of a single production run? Do all units have the same kind of lubrication as well as the same amount? Did you test to see if the tension on the front truck screws was the same for every unit? How about the resilience of the springs on those pilot trucks? Were all of the locomotives broken in to the same degree? Did they all have the same amount of running time? And what about the track? Did you institute any systematic differences when you assembled, disassembled and reassembled the track? Did each circle you constructed have exactly equally tight joints and alignment? What about the coefficients of friction for the rails and wheels; were they all the same? There are many more.

You stated that all specifications which you measured were within 1/16 inch. You do realize that's the equivalent of nearly five-and-a-half-inches on the prototype. I have to say, I'm not impressed with measurement which are as great as 1/16 inch on something as small as an HO model, particularly when we are discussing running gear and its performance.

It's a fundamental statistical error to predict beyond one's sample unless that sample was appropriately constructed. Your sample is so small and so uncontrolled that you really can't generalize from it at all. Yes, I kind of did so when I was explaining the gambler's fallacy but I wasn't asserting any meaning beyond your findings.

You have another problem, too. One does not run the study, do some statistical analysis and then go to a statistics book to see what confidence limits were met. That is just a terrible misuse of the results. The way it is supposed to be done is one develops an experimental design with the expectation that there will be no differences (the null hypothesis), one recruits a carefully selected sample of a size that enables one to extrapolate beyond the sample findings at a degree of certainly which is determined a priori, the subjects are run, and a statistical analysis is completed to see if the null hypothesis can be rejected. Do I expect you or anyone else on this board or at Bachmann's to do this kind of rigorous testing? Of course not. But I'm not comfortable seeing someone present data as if it were appropriately derived when it clearly was not.

All of that said, I wouldn't be at all surprised to find that your 30% won't go around 18-inch curves is pretty close to the truth. And if it is, I believe it would behoove Bachmann to look at their engineering to see why some Connies can handle tight curves while others cannot.
                                                                                                                                                                                -- D

[ACY]

Don, I must say, you really know your statistics. And like I said before with the circumstances I did the best I could with what I had to work with. Some of the things you mentioned were controlled for bit many were not and of course many variables were unknown. But it would be extremely difficult for the average person to do much better than I did. To conduct a statistical analysis in the manner you suggested would be next to impossible unfortunately. That is why I had to make a few assumptions about unknowns and also work hard to try and extrapolare any information from the data recorded.

To all, I said all along I doubt it affects 30%, but I would not rule out it affecting 5%-10% of the production run.
One piece of information I find interesting is that my friend who was with Lionel would test about a hundred locomotives a day on a set of 8 patallel tracks and begins on the most gradual radius and gets progressively sharper on the inside, he would take note of what the sharpest radius the locomotive could negotiate was. For a certain O scale Lionel locomotive, (I can't recall which anymore), he observed that roughly 90% could negotiate O-27 while the other 10% could not. His sample size was a few thousand locomotives. So even though my sample size was not nearly as large, the findings are in the ball park of a much larger study involving Lionel O scale locomotives. He said the issue was most likely related to the mass production of the locomotive in a factory with less than desirable QC, however the factory is still used because it has lower costs than otger factories with more QC and consistency throughtoutthe whole production run. In fact, Lionel produces model Nascar cars in two different factories, the model cars are marked slightly differently on the bottom but are otherwise identical. However, the cars marked from the one factory tend to have errors in the decal placement orthe paint or other details while cars from the other factory are without any noticeable flaws.

So it could be ppossible Bachmann utilized multiple different factories or tooling which caused minor variations in the ptoduction quality. Regardless I still find Bachmann to be the most reliable and detailed locomotives and at an excellent price point. Bachmann locomotives are a good reliable product especially considering how many units they have to make of each locomotive.

[jbrock27]

my main vocational training as a clinical psychologist

D, I know, this is why I refer to you as "Doc".  Another impressive item on your resume, or curriculum vitae if you will.  I don't doubt for one second you are more qualified to discuss statistics.

Comparing O Scale to HO Scale?  Sounds like comparing apples to oranges to me.   But what do I know?  I don't have any data to ramble on about, just what I see in operation.

[Doneldon]

ACY-

I understand that it really isn't feasible to do a fully controlled, properly constructed experiment to determine the proportion of
Bachmann locos which can negotiate 18" curves, or any other characteristic for that matter. I was just making the point that we shouldn't
accept what data we do have, like your informal little survey, as though it were developed with the kind of rigor needed to make such
predictions.
                     -- D

[rogertra]

Last time I used an 18" radius curve on a running line was on my teenage Hornby Dublo railway.

Since then it's been either hand laid or this time around, flex track.  Smallest radius 30" on running lines.

However, my old GER did have a 20" (roughly) radius leg on the wye in Berger Yard that was used for turning engines. In view of the 20" radius on the wye, the largest engines used on the branch up to Berger Yard were 2-8-2 and as the 20" radius leg was the farthest from the aisle, the 2-8-2s looked acceptable from the aisle viewing angle.  My current GER has an 18", roughly, curve in the industrial area serving as the lead to the J. King & Co. warehouse.  This curve is currently switched by an 0-8-0 but will eventually be switched by an 0-6-0, once the 0-6-0 is equipped with a DCC sound decoder.  See curve here, scenery still nowhere near completed: -



I realise that some people have space issues but I would still try to avoid 18" curves on running lines and shoot for 24", just because medium sized equipment just doesn't look "right" negotiating curves as tight as 18" and larger engines and passenger cars look, well, really "odd".

Cheers.

Roger T.

[ACY]

ACY-
I understand that it really isn't feasible to do a fully controlled, properly constructed experiment to determine the proportion of
Bachmann locos which can negotiate 18" curves, or any other characteristic for that matter. I was just making the point that we shouldn't
accept what data we do have, like your informal little survey, as though it were developed with the kind of rigor needed to make such
predictions.

I understand your view on my findings, but what we do know is some percentage of many HO locomotives will not be able to negotiate 18" radius curves even though the majority are capable. And the only cause in many cases is due to variation present in mass production runs. I still think it is reasonable to believe that some small percentage such as 1% may be affected. But beyond that I probably shouldn't speculate. However I am working with a friend who may be able to facilitate a larger scale analysis of around 100 identical locomotives but even that may not be sufficient unfortunately. So I am holding off for now until I figure out what sample size would be necessary to perform a statistical analysis that holds more weight so to speak.

[Rielag]

Thanks to all,
    The info contained in these replies  has convinced me. Going to a combo of 22 and 18 and stick with the 2-6-0  steam and the four axel diesels until I can shake loose more space in the basement. For now 22 outside and 18 inside(two ovals) on a 4x8 platform.

                                                                                                 thanks Gene