q question about curve fitting JPL coordinates

I am NOT a mathematician. I cant do arithmetic. so if I seem stupid
please just correct me.

I am a programmer and have TURBO 7.0 compiler. I also have a book with
lots of pascal code in it. I compiled up the cross program " SQUARE,
it will cross large arrays.

After I cross (a Ta) it returns an equation then using GaussJordan
algorithm I solve the equation.

My input is the X Y Z coordinates from JPL. I I modify X2+Z2/2 and Y2+Z2/2

The purpose is to simplify the fit to the program. It expects X, an Y only.

I get back TWO values (A and B) to fit the equation 12/A2 +Y2/B2 = 1

my residuals are taken from this form

I wrote a simple plot and surprise,PLUTO,NEPTUNE,URANUS,SATURN,Jupiter
and MARS residual calculations all show these sine wave shapes.

I first noticed this with 1981 to 1983. URANUS showed an increasing
amplitude. like a sound wave increasing in amplitude from left to right.

I fitted and plotted the differentials form 1981 to 2007 and JUPITER
shows 4 sine peaks. Mars shows a similar pattern but the points appear
to osculate from one side to the other.

My basic question is, Why do I see these artifacts in JPL DATA. I can
run the code to produce XY&Z. But JPL (in the astronomical journal)
provides X Y Z dx dy dz for all major planets.

has anyone ever done this CURVEFIT the published XYZ data. is this a
known problem. And I know the the coordinates are good enough to
generate ephemeris. I was NOT trying to find the ELEMENTS, I just was
curious I expected small errors, not values from ± from .001 to .00001,
in a sine pattern, The errors for moons are much smaller than these
values.

No I have not looked at perturbations. The sine waves I see all seem to
be centered on the data set for each planet. and the different
functions all indicate artifacts not program errors. I am only looking
at superior planets. they have periods greater than 200 days.

josephus