# precision: how many measurements? - (Dec/31/2005 )

Hello,

I have designed a new type of volumeter, an instrument used in my research to measure arm volume in patients who have complications after surgery for breast cancer, namely arm swelling.

In order to validate my instrument, and compare it to the one used in most hospitals, I first want to take objects of a known volume (which I shall call reference volumes), and compare the known volumes to the measured volumes both with the standard volumeter and to my own device.

I want to look at 2 facets: accuracy (the difference between the known standard volume and the volume measured with both the standard volumeter and my own) and precision (the variability of the measured volumes with each instrument).

I have about 10 reference volumes and I will end up with 2 curves (calculated vs measured volumes), one for the standard volumeter, and one for my own.

Question: for each reference volumes, how many measurements should I make in order to get a statistically valid appreciation of precision?

Thank you very much and cordial regards to all,

Jean

-Jean-

QUOTE (Jean @ Dec 31 2005, 04:43 PM)
Hello,

I have designed a new type of volumeter, an instrument used in my research to measure arm volume in patients who have complications after surgery for breast cancer, namely arm swelling.

In order to validate my instrument, and compare it to the one used in most hospitals, I first want to take objects of a known volume (which I shall call reference volumes), and compare the known volumes to the measured volumes both with the standard volumeter and to my own device.

I want to look at 2 facets: accuracy (the difference between the known standard volume and the volume measured with both the standard volumeter and my own) and precision (the variability of the measured volumes with each instrument).

I have about 10 reference volumes and I will end up with 2 curves (calculated vs measured volumes), one for the standard volumeter, and one for my own.

Question: for each reference volumes, how many measurements should I make in order to get a statistically valid appreciation of precision?

Thank you very much and cordial regards to all,

Jean

A safe rule of thumb is 30.

-labmath-

we usually use 25 as a rule of thumb..... but in these things, its really more the better!

-janbrisbane-