How many digits should be used when measuring density?

[LTobiasL]

I'm measuring density. The equipment shows 3 digits after the point (1,481).

When I perform a MSA with the 3 digits the values will be over 10%. But when only using 2 digits (1,48 and so on) the value is below 10 %. (The tolerance is 0,02).

What's the conclusion. The system is ok but the measurement results should only be showed with 2 digits?

Thanks!

 

[Duke Okes]

Typically measurements should be recorded at one more decimal place than the specification has.

 

[Miner]

You should attach your data for a definite answer. One probability that comes to mind is the effect of inadequate gage resolution. Inadequate gage resolution can dramatically affect the %GRR number moving it up or down. The %GRR obtained is not valid unless the resolution is acceptable.

 

[DsqrdDGD909]

I'm measuring density. The equipment shows 3 digits after the point (1,481).

When I perform a MSA with the 3 digits the values will be over 10%. But when only using 2 digits (1,48 and so on) the value is below 10 %. (The tolerance is 0,02).

What's the conclusion. The system is ok but the measurement results should only be showed with 2 digits?

Thanks!

How are you measuring it? We do densities to 3 decimal places. Our specs are to 2 decimal places.

 

[Dave Dunn]

I'm measuring density. The equipment shows 3 digits after the point (1,481).

When I perform a MSA with the 3 digits the values will be over 10%. But when only using 2 digits (1,48 and so on) the value is below 10 %. (The tolerance is 0,02).

What's the conclusion. The system is ok but the measurement results should only be showed with 2 digits?

Thanks!

If your tolerance is 0.02 total but you're rounding everything to the second decimal place, you only have at best 3 distinct readings that fall within your tolerance. By doing this you artificially reduce observed variation.

By analogy, if you needed to measure a length with a tolerance of ±1 inch, but you rounded everything to the closest inch you would have very little observed variation in your measurements, even though the actual lengths could vary by nearly 1/2" large or small. Lengths of 1", 1.49", .501" would all round to 1" and zero variance. Looks good on paper, but doesn't give you a true picture of the parts.

In short, by excessive rounding, your system is no longer acceptable and GR&R testing would not be valid.

 

[Randy]

2 digits, normally the thumb and forefinger