Smoothing data, 'rlowess' behaviour.

Hi all,

I was just wondering whether the 'rlowess' smoothing behaviour that I have plotted below makes sense (red and blue lines), as at the end of the graph, even though the scatter points are more frequent at smaller values, the plotted line increases in values, which is rather counter intuitive.

What i have done is that i have smoothed my scatter data using the smooth function, 'rlowess' option, and then on the smoothed function's returned values i have used the fit function with the 'fourier8' option to further smooth out the data and be able to extract values at any point on the plot.

The same behaviour (and more intense) appears if I use the 'moving' option for the smoothing of the data, and pretty much noting changes if I use a 'poly9' option.

So what is the statistic explanation -if you will- of the behaviour of the blue and red lines as displayed below?

Thanks in advance for your responses & help.

KMT

P.S. I was also wondering why the 'rlowess' won't go on to plot until the end of the scatter set, which goes roughly up to 76 units on the x-axis, yet the red and blue lines stop at roughly 60 units on the x-axis

[Image Analyst]

Are those points all for the same engine? So like at a load of 50% the heat could be anything from .75 to 3.2? Why is it so variable? Are the engine variable, or is it your measurement technique that is causing the variability? Do you trust your measurements?