Hello All:
We prepare our own in-house reference material for Au analysis (used
in conjunction with certified materials). About 200 lbs of material
is collected, sieved, mixed and bottled. I would like to test this
material for homogeneity. We have an old spreadsheet with a ANOVA
test where a sample from the top and bottom of each bottle is
analysed and input a a pair. Unfortunatley the spreadsheet was not
locked and over time has become corrupt. I have had some basic stats
but it's been too long. Most of the ANOVA tests I've found (purusing
my old texts) are for 3 or more groups. This sheet appears to calc
the mean and sd of each pair row-wise, along with some calcs like ni
and ni-1. A grand mean and sd are also calc and there are some sums
along the bottom. The whole thing culminates in a ratio which gives
F calc and we look up F critical. If F calc is less than the F crit,
the mixture is considered homogeneous at a certain level. I have
tried to resurrect the spreadsheet but am uncertain if I am correct.
Can anyone shed some light on this matter, point me in the right
direction, or point me towards a standard data set with known results
that we could check our calculations against.. Is this the correct
way to do this or should we be looking at some other test(s) (say
comparing the means and then comparing the variances). Picking the
correct test was always the trick for me.
I also have a question.....I don't quite get why, in the F table,
the ratios get larger as the level (is it confidence level?) gets
higher. The current batch we are working on calulates to an F of
alomst 5 and the critical level at 95% is about 3 but id you look at
the table, the critical number get higher as the level goes up. I
would have thought it would be visa versa. Under this scenerio we
could just pick a higher level and our numbers would pass.
Any help is much appreciated. Thank you in advance!
Wishing I had paid more attention in Stats class,
Terry
|