Adjusting outliers

Well, it had to happen at some point: a usually reliable model requires multiple repair trips for one owner. In statistics, this is known as an “outlier.” The question for me: how to handle them? Leaving outliers as they are can distort the results, especially with small sample sizes. In the most extreme case from the upcoming August results, I… More →

Why TrueDelta excludes inconsistent respondents from the analysis

TrueDelta only recently started collecting data on the 2007 Acura MDX. So far, the 32 owners who have responded have reported only one trip to the shop that included a successful repair. This repair trip would be included in the analysis, except for one thing: the repair was reported for January, and the owner in question did not respond for… More →

Issue tracking added to the Vehicle Reliability Survey

Another quarter, another set of tweaks to the Vehicle Reliability Survey. This time around, I’ve sought to address two common problems with responses. First off, people often report multiple issues at the same time. The form has been revised to make it clearer that the top half of the form should be filled out separately for each repair trip, while… More →

Long-term road tests: insufficient sample

For years the car magazines have conducted long-term road tests (links from Google). The manufacturer gives them a car to drive for 30,000 to 40,000 miles, and they write up both their impressions of what the car is like to drive and what broke. No problem with the first bit: you might learn more about a car when you drive… More →