In Article  <9k0r5t$jvm$1@news1.xs4all.nl> Josh wrote:
> That was about the trend, but the total average seems
> fairly established at .5 seconds slower per day.

Indeed, on our Troubled Times site (or pending in the queue) are the
following:

1. a finding that per the Earth Rotation Service and
   other official databanks, the Earth was slowing in
   its rotation some 54.87 seconds/year in 1986, and
   is now up to 64.0908 seconds/year (previous post)

2. a finding that the atomic clock has been adjusted WAY
   past the claimed inserted leap seconds the public is
   being led to believe are the only manipulations done
   to clocks by the clock-masters, the US Navy (below).

Note: related chart 

From the Troubled Times Slowing TOPIC pending queue.

@@@@@@@@@@@@@@@@@@@@@@@@@@@@

Comparison of our clocks to Atomic time results in an average of +40
sec/year for each the last 3 years (Jul 98 - Jul 01).  This is added
time making our clocks appear to run faster and faster each year.  Josx
is measuring actual earth rotation time and will publish his results
once something definite can be  determined.  He has already indicated
that the earth is rotating slower by .54 sec/day than his expensive
accurate stopwatch.  Assuming this watch was in sink with Atomic Time
with in plus or minus .34 sec at the factory, then as he indicated he
could already be measuring a slowing trend.

Mathematics of Atomic Time comparison measurements

Measure Drift "R" as the difference of Atomic time and individual Clock
time (approximately 3 months intervals) = R1, R2, R3…. RN.  Assumes
clocks set back to atomic time beginning of each period otherwise
determine a difference from last time for RN. (+ = Clock reading faster
than Atomic Time.  - = Clock reading slower than Atomic time)

Change in drift then = R2 - R1,  R3 - R2, ... RN - RN-1

Change in drift/day (CN) =  (R2 - R1)/ D1,  (R3 - R2)/ D2, ... (RN -
RN-1)/DN-1
Where DN-1 = Number of days (calculated to 5 decimal places) between
measurements.

Next select the most reliable clocks using calculated standard deviation
as a reference. Average all "change in drift/day (CN)" readings for all
selected clocks during each time period  or AvgCN .

Average driFt curve AvgFN = first Integral of AvgCN per each time of
measurement  =  (AvgC1)* D1, (AvgC1+AvgC2)* D2, (AvgC1+AvgC2+AvgC3)*D3,
...  AvgFN

Average comparison results to Atomic time = Integral of AvgDFN per each
time of measurement =  (AvgF1), (AvgF1+AvgF2), (AvgF1+AvgF2+AvgF3),  ...

This plotted becomes the 2nd integral results.

Note that all straight line trends in the original data are filtered
out, as desired.  Only the changing situation is plotted in the results.