mkaresh

I wanted more up-to-date car reliability information that included actual repair rates. So a few years ago I started getting people together to make this possible. TrueDelta now updates actual repair frequencies, not just dots, four times a year, to track cars closely as they age.

We recently updated reliability stats for the Corvette, based on owner experiences through June 30, 2014. In terms of repair trips per 100 cars per year:

2014 Corvette: 43, moderate

Noteworthy: of the 26 cars in the sample, only two required a repair in the survey period. If anything, the fairly small sample size is probably overestimating the repair frequency. Even as reported it's not bad for a first model year car.

We'll have updates in November and February. For repair descriptions, the stats of other cars, and to sign up to help improve this information:

Chevrolet Corvette reliability ratings and comparisons

-CM-

From the site: Chart based on 24 repairs.

I don't think any meaningful conclusions can be drawn from such a low sampling.

mkaresh

It's not a large enough sample to draw conclusions about the frequency of specific repairs.

The sample size was sufficient for an overall repair frequency, where the relevant number is the number of cars surveyed. This was 26 in the previous round and, as noted on the site (hover over stats for notes) 24 of these 26 reported no repairs during the survey period--I meant to mention this in the OP. The rest of the repairs were before the survey period, after it (will be included in the next update), involved only software updates, or did not fix the problem.

From this I'd personally conclude that few of these cars have required repairs so far.

Not that I'm against larger sample sizes. I look forward to having more owners participate.

KenHorse

Yup. Statistically insignificant

Zymurgy

Sample size is insufficient for any purpose. To have a 95% confidence level with a 5% confidence interval, the sample size would have to be 380 of the 37,288 2014 Corvettes built.

Your sample size (26) gives a 95% confidence level with a +/- 19% confidence interval. LOL

mksz51

Yes Sir - we have a statistician in the house!!!!

mkaresh

Not disputing your math, though with a large population variance rather than population size is the best basis for figuring confidence intervals.

But we're not conducting medical research here. Why is it necessary to have a 5% confidence interval? Maybe the actual repair frequency is 33. Or maybe it's 53. I'd classify either as "moderate." It's not near zero, where the most reliable models are. And it's nowhere near 100, where the least reliable models are.

I have the advantage of having another quarter of data in nearly clean form. With a few more cars involved, and nearly twice as many car-months covered, it looks like the stat will barely be changing in the next update. This stability in the stats from quarter to quarter is common, and isn't what would be expected if they were essentially shots in the dark.

Another way of looking at the stat is to put it in terms of real-life experience. If you asked 30 owners if they'd had any repairs, and 26 said they'd had none, would you infer anything from this?

All of this said, I would also prefer larger sample sizes, and hope to have these with future quarterly updates.

Zymurgy

A 5% confidence interval is pretty much the standard in any statistical sampling. Sure, you can have higher confidence intervals, but it means that your conclusion has a much greater chance of being incorrect. Saying that the repair frequency is 43% +/- 5% (at the 95% confidence level) means that there is a 95% likelihood of the actual number being between 38% and 48%. With 43% +/- 19%, the actual number probably lies somewhere between 24% and 62%. One of these I would trust, one I would not.

HogwildC7

What about value? Just took mines to a Caddy dealer and was told that my 2015 Z51 LT3, with all options. MSRP $77500, was just told that with 4800 miles it is worth $62K. So why are they priced so high?

Steve_R

Winner! You saved me from doing the same calcs. Even without calcs common sense says a sample size of 26 from a population of over 37,000 is meaningless.

ckwhite

mkaresh

Common sense is often out of line with statistical methodology. Once a population is fairly large, the size of the population doesn't have a bearing on the required sample size. This is how polls can be accurate when sampling a few hundred people out of a population of millions.

Again, I'm not disagreeing that a larger sample size wouldn't be better, only that "meaningless" is a bit strong.

The average for a 2014 model in this survey is 26 repair trips per 100 cars per year. Taking the above calculations as correct, there's a slightly better than 5% chance that the 2014 Corvette's reliability is better than average, and a much higher chance that it's worse than average, if probably not much worse. Personally, while I'd like to be able to offer more precise stats, I don't think this inference is meaningless.

Corgidog1

Too many variables---ie if one set of 30 owners were from urban cities and another set of 30 owners were from rural areas, one set of 30 owners were from a cold climate and another set of 30 owners were from a very warm climate, cars that are tracked vs. cars not tracked, etc. the results might differ. Technically a statistically valid stratified sample might be more appropiate in this case to handle the various variables.

WelderGuy

My present car scores 0, and that's also the number of problems I've had since I bought it five years ago.

NavyBlue2

Wow!
A 5 yr old C7!!!

gixxerbill

Dang Zymurgy I'm impressed. You actually took notes in stat class.

mkaresh

Paradoxically, discussing the sample size tends to have the effect of slowing the growth of said sample size.

One thing to realize: if a difference is so small that it can't be effectively measured with 30 cars, it's not large enough to matter when buying one car.

This is one of the key lessons from our stats: many of the differences aren't large enough that people should buy one car rather than another because of them. Other reliability information, which disguises the size of the differences behind dots, lead people to think these differences are larger than they actually are.

mkaresh

So, we've updated our stat based on a few more cars (total now 31) and three more months of data (covering through the end of September)...and it's virtually unchanged, 44 repair trips per 100 cars per year. This is a moderate repair frequency, a little worse than average.

Reported problems mostly involve panel fits or issues with the infotainment system. When the latter were addressed with free software updates, we didn't count them as repairs.

As discussed earlier, the sample size is fairly small. We should have a larger sample size with the next update, in February. How much larger is up to owners.

To view the repairs behind these numbers, check the stats for other cars, and sign up to help improve this information:

http://www.truedelta.com/Chevrolet-C...reliability-52

BIC

Although there are obviously statistical concerns with what you are doing, I'm OK with it as long as you keep the reader informed of the issues (as others are doing here). So, I would like to thank you for your efforts in this endevour and hopefully, it will become more statistically valid in the future. I look forward to it.

dvilin

I will just add nice attempt, but, well you know the rest.