Amsoil AFL 5W40, 5807 miles, 2008 VW GLI 2.0T

[RI_RS4]

pickled,

rhouse181's graph is meaningless, since it has not taken into account the normal drop in metal due to wear-in, or residual contamination from the residue left from the previous oil. Those numbers at the low side of the graph are heavily biased higher, because of residuals and the fact that short OCI's are common (and/or required) in the early life of an engine. This explains the curve shape, and the high variance from the fit.

 

[pickled]

There is meaning to all data...hence my interpretation that his graph is a poor predictor of ppm .

 

[rhouse181]

I agree that there is much left to be desired with sample group of data I had access too. There is plenty of truth to your concerns with the validity of the trend. There are only 30 observation points in the sample. Also, there aren't a lot of higher mileage engines being sampled because most cars a max of 2-3 years old...

However, I strongly disagree that the chart is meaningless. Sure, you could screen the data left and right to eliminate certain parameters. But we do not have access to such statistically perfect data. There are a huge amount of outside factors influcing each and every oil sample. Any one of these "statistical variances" would blow a perfect correlation out of the water.

With only 30 observations points of our non-perfect/real world data, an r^2 value of over .5 certainly shows a positive correlation between the two variables.

None of this is ground breaking... i don't think one of us would be surprises that iron has some positive correlation to mileage on oil. Just thought I would share an "intriquing" trend to provide a linear representation to the data...

 

[RI_RS4]

Never mind. I give up.

 

[VeeDubb]

Originally Posted By: rhouse181



With only 30 observations points of our non-perfect/real world data, an r^2 value of over .5 certainly shows a positive correlation between the two variables.

None of this is ground breaking... i don't think one of us would be surprises that iron has some positive correlation to mileage on oil. Just thought I would share an "intriquing" trend to provide a linear representation to the data...



rhouse, thanks for providing the data and analysis. But I have to agree with RS4. I don't think RS4 is arguing that there isn't a correlation. I think what he is saying is that your nonlinear graph may be biased given that you don't control for a couple of variables. So it would be a stretch for you to make statements such as:

"You are pretty much on the trendline for iron ppm at 5k miles. Iron wear definitely looks normal.....Seems like 5k is a good change interval."


But it's easy to fix it. Just include vehicle miles as an additional regressor and a dummy variable that takes value of "1" if it is the first run on the oil. Rerun the regression and your coefficient on the miles on oil variable will be a partial correlation that nets out the influence of vehicle mile and 1st oil run effects. Adding two variables will only cost you two degrees of freedom so your small sample issue is not that big of a deal.

 

[ADFD1]

Originally Posted By: saaber1

How do you define "abnormal"? I think throwing 150, 300, 350, whatever ppm numbers out there is pretty much meaningless. We have 39 UOAs on this engine and we know what "normal" ppm is for this engine at least for that data set. The OP's iron is 6.4 ppm/1000 miles at about 24k miles. That is just below the norm. Sorry this chart hasn't been updated but here it is.


To revisit the other threads about "are UOA metals an indication of actual wear?", I would summarize three posts that are very illustrative to me:

(1) One describes a real-world example where teardown showed large amounts of cam wear and the corresponding UOA sequence showed only a rise from 15ppm to 18ppm. In other words 15ppm was normal for that vehicle and it rose to only 18ppm when the wear problem occurred.

(2) another post describes catastrophic failure of the cam follower on the same engine as the OPs and it showed only 25 ppm during that time which is well within "normal"

(3) Buster's post from the redline guy that showed UOAs don't pick up the large particles and can show higher ppm due chemical "wear" which does not mean the engine is wearing faster

As countless others have said, contamination or wear metals can be helpful to show if you are out of the "norm" and should investigate things farther. But IMO putting "arbitrary" numbers such as 500ppm is really meaningless.



Three reasons to me why unless I feel I have a coolant leak or bad air intake I wouldn't spend any money on a UOA. 3ppm rise in iron and tear down shows the cam is shot. 25ppm normal, and the engine suffers catastrophic failure. No thanks.