Got a new scale so I could more easily read the numbers. The new one is digital.
At the low end, they read nearly the same. The old one and the new one with steel weights read 35lbs of steel weights at 35 (old) and the new one 34.8.
At my weight they read 5 lbs difference - the new one reading heavier. I don’t know if I can stack enough steel weights on the scales to determine the accuracy at an upper limit.
How ideas how I might I determine which is the correct weight?
You don’t say what type of scale, but you should at least say what type of accuracy you have to obtain and the type of scale. I had to deal with scales and calibration for work. I finally convinced them to hire a calibration service because the cost of all the weights needed to insure calibration across the range of the machine and the weight standards calibration costs was too much.
Load cell electronic scales have a non linear response curve. Some can be calibrated with more than one or two weights so that the electronics can adjust for this curve. It sounds like you may be dealing with an electronic scale.
Well, you can try weighing yourself on both and then weighing yourself on both holding 35lbs of steel weights. If only one of the scales is wrong, you can estimate which one and approximate by how much by building the following table:
Scale A | Scale B
Steel Weight 35 34.8
You X X + 5
You + Weight X+Y X+5+Z
This isn’t a definitive test, but if Z happens to be ~35 lbs then most likely your old scale is wrong. If Y happens to be 35lbs then most likely your new scale is wrong. If both Z and Y are correct, you might want to call an exorcist 
The new scale is Taylor 7014B Lithium Digital Scale, the old one is some generic spring loaded one from Walmart or something.
Following groman’s advice, I just weighed myself on both with the weights in hand and they are consistently 5 lbs off at the top end.
I guess in the end, I’m more interested in the amount I have lost, so I guess it won’t really make a big difference, but it would be nice to know a true amount.
A man with one watch knows what time it is, a man with two is never sure.
Is this what you have?
Overall, I would trust the Taylor over the spring scale, with a few caveats:
Make sure your scale is resting evenly on a flat, hard surface. Soft surfaces (carpet) may effect your readings.
Make sure you have fresh batteries in it. Low or inconsistent excitation voltage to the measuring cell can affect accuracy, although I would think that if the battery is strong enough to power up the digital indicator, it should be strong enough to provide accurate loadcell response.
Make sure their isn’t some fiddley-bit stuck in your scale, such as a piece of “packing” or a travel-stop inserted at the factory.
Harmonious Discord: modern electronic cells should have a linear response curve out-of-box. In application, foreign debris, and connection of various devices (mixer motors, agitators) and plumbing fixtures (pipes, valves, and control mechanisms to such) to the weighing device (like a tank scale, or a small tank sitting on a floor scale) are, IME, the major causes of non-linear response.
One of our customers makes plastic bottles for things like laundry detergent and spray cleaners. They called us out yesterday because one of their floor scales was weighing erratically.
I lifted up the scale, and there was a big blob of melted plastic wedged underneath it. We often repair customer’s scales by swiping a feeler gauge through the open part of single-ended beam loadcells used in most floor-scale applications; this pulls out “gunk” effecting the loadcell’s response.
FTR, I’m a licensed scale tech in two states. I mostly deal with large-scale industrial apps., counting apps., and legal-for-trade, but I do have some experience with analyticals, too.
If a small application scale like an analytical, counting/retail scale, or a “bathroom scale” needs a linear calibration, there’s either something stuck in it, or it has a bad loadcell, or it’s a bad design.
FWIW I got curious about the behavior of my digital bathroom scale and did some experimenting, and convinced myself that it remembers the most recent reading, and if the next reading would be very close to it, the scale alters the reading to match that most recent one.
My scale would reproduce my weight to its 0.2 pound resolution every time I stood on it, if I kept trying over and over again. However, if I hold a small weight, like a liter bottle of water weighing about 2 pounds, and weigh myself, I get a different reading (which I expect), and if I then weigh myself without the bottle, it will usually not match the last bottle-free reading. The new reading will then keep repeating with subsequent readings unless I break the series by holding another weight.
My guess is that they pulled this trick to eliminate constant consumer complaints that the last digit changes between retests. I always want to have at least a couple digits that aren’t fully meaningful on a display, but I think many consumers expect ALL the digits to repeat (that is, though they don’t think of it this way, most consumers want the partially significant digits hidden).
I am properly convinced that the digital scale is correct. :mad:
However, since the difference between the two remains the same at the high end, I can still claim that I lost 17 lbs.