Previous in Forum: What is the Trend of Global Abrasive Business?   Next in Forum: Socket Weld VS Butt Weld (Joining or Fittings)
Close
Close
Close
2 comments
Associate

Join Date: Mar 2013
Location: North East UK
Posts: 27
Good Answers: 2

Osmolality Testing of Liquids

03/13/2013 10:27 AM

Hi Everyone,

quick question, we test liquids for osmolality using the freeze point depression method and when looking at our test method it states we do it twice and take the average but when I have looked at the EP and USP Pharmacopoeia it doesnt state to do it twice so does anyone know if this is just something my site has done historically or is there a basis for this in a document I havent managed to find?

Thanks

Register to Reply
Pathfinder Tags: Osmolality testing
Interested in this topic? By joining CR4 you can "subscribe" to
this discussion and receive notification when new comments are added.
Guru

Join Date: Aug 2008
Location: England & Ireland
Posts: 1063
Good Answers: 60
#1

Re: Osmolality Testing of Liquids

03/13/2013 11:52 AM

It is more normal to take a measurement 3 times before averaging. I'm sure someone well versed in statistics could tell you why. Just intuitively, one measurement could be a fluke. Two flukes in succession is very unlikely, but if one is a fluke, then the average of the two is obviously biased. Therefore three measurements is the bare minimum to make before averaging. It's certainly what they use in the measurement of elemental composition in analytical chemistry.

Other more tedious methods would be to take 10 measurements, chuck out the highest and lowest, then average the rest.

Register to Reply
Guru
Safety - Hazmat - New Member United States - US - Statue of Liberty - New Member Engineering Fields - Chemical Engineering - Old Hand

Join Date: Mar 2011
Location: Lubbock, Texas
Posts: 14331
Good Answers: 161
#2

Re: Osmolality Testing of Liquids

03/14/2013 8:49 AM

For any scientific measurement to have a sound basis statistically, there should a minimum of three measurements (simply because some statistical parameters become essentially meaningless when the data set is so limited). Significance testing requires four results for a reliable Student's t test of confidence limits. Anything more than about 20 tests there is little to be gained in understanding the statistical distribution of data.

__________________
If it ain't broke, don't fix it. Just build a better one.
Register to Reply
Register to Reply 2 comments
Interested in this topic? By joining CR4 you can "subscribe" to
this discussion and receive notification when new comments are added.

Previous in Forum: What is the Trend of Global Abrasive Business?   Next in Forum: Socket Weld VS Butt Weld (Joining or Fittings)

Advertisement