| mkaresh |
A thread about TrueDelta's research is stickied at the top of this forum. I just wanted to give those who've been helping out a heads-up on the initial results.
So far I have 68 months of data from 32 owners, which is a bit below the minimums for making the results "official." So they'll be asterisked and visible only to panel members--except on this forum.
And the results are...no new MDX owner has reported a repair yet that meets the criteria for being included in the analysis. Five repair trips have been reported, but all for things that I exclude for one reason or another: damage was caused by the dealer, dealer denied there was an issue, etc.
Actually, one person reported a frozen door lock, but they didn't check in at the end of the quarter, so their data isn't included in the analysis. Continuous participation is required so people don't just respond when they have problems to report, which would skew the results.
So, while the sample size so far is small, the early indication is that the MDX will be very reliable. Time will tell, and there will be updates every three months. There will almost certainly be enough data for the result to be "official" next time, in August.
The repair rate will of course not remain zero with more data. But a low repair rate of about 0.3 repair trips per year seems likely based on general trends I've been observing in the data. This number would be typical of Hondas and Toyotas not in their first model year, and so quite good for a model in its first year.
Those of you who have been helping out, thanks, I appreciate it. |
|
|
| oilchange |
quote:
Continuous participation is required so people don't just respond when they have problems to report, which would skew the results
Requiring follow up causes a systematic error underestimating problems. Owners can have a problem fixed in one visit that was significant. Some owners do not report back in a case of a real problem for a number of reasons such as it is fixed and don't look back, it is not fixed by that stupid dealer and can't trust the dealer to do anything right-can't be bothered to really get if fixed, too busy, and so on. Systematic error and small numbers make for statistically not valid data and skew the results. |
|
|
| mkaresh |
Generally the criticism I hear is that I'm ONLY going to hear from people when they have a problem, which would distort problem rates upward. So I designed a process to eliminate this problem. Other reliability research has no such control, and so is susceptible to this source of distortion.
Now I'm hearing that refusing to let such owners have an impact will distort problem rates downward.
Sorry, but it just isn't so. The most valid results will come from relying on those people who consistently respond quarter after quarter. These people report in whether or not they have a problem.
Excluding those that do not consistently respond reduces the sample size, but does not distort the results. Luckily I'm finding that most people who respond do so consistently.
Let's say we have 200 owners. 100 respond consistently, while 100 respond only when they have a problem. Let's say that 10 in each group have a repair.
Including all 200 owners in the analysis would make the estimated problem rate 16.67% (20 problems divided by 120 responses), when the actual is 10%. Including only the 100 consistent responders would yield an accurate result.
The small sample sizes are currently an issue, but one that will be going away. |
|
|
|
