hi,
what are your research questions?
Hi guys help me? I have a problem. I have to analyze a product just finished the processing (hot) and subsequently when it is cooled to room temperature. These are the results of a year of analysis. What do you consider statistical tool I can use to not make the first test, then only the second at room temperature?
HOT AMBIENTAL
A 102,8 101,4
B 100,7 98,9
C 101,6 102,3
D 102,6 99,0
A 102,5 104,3
B 99,0 101,5
C 99,6 100,8
D 99,6 100,4
A 100,0 98,0
B 103,5 103,3
C 103,1 101,7
D 99,8 99,7
A 102,7 102,9
B 104,3 102,7
C 103,1 101,7
D 99,8 99,7
A 102,7 102,9
B 104,3 102,7
C 102,9 100,1
hi,
what are your research questions?
my target is to reduce the number of chemical analysis. Instead of doing the analysis before and after, I would make them only "after".
I am looking for a statistical rationale to be authorized
Hi,
so you would need to prove that you can predict (actually postdict ) the "before" value from the "after", right? I would look at a Bland-Altmann plot first and depending on the result a regression or a test of the differences against a constant. Is what you showed all the data you have or just an example?
regards
I have a specific limit that I have to respect for each batches (the values must be between 95 and 105).
So they are all batches comply (before and after).
These are the values of an entire year (I have only these)
Why do you do both analyses as of today? Is the limit defined for the before or the after measurement? And what are the A, B, C, Ds, different products?
I do both analyzes because the customer wants it.
Limits are the same for both of the steps
A, B, C, are the same product (product in different days)
you are definitely right. I explain to you:
HOT = is the sample in the tanks
ROOM TEMPERATURE = the sample in the vials (ampoule)
I would like to explain to the customer that it is useless to perform test in the tank, because in the test vial always conforms (and "similar")
Last edited by crimelist; 03-10-2016 at 07:49 AM.
hi,
following the logic of the Bland-Altman plot you could look at the relationship between the average of hot and ambient measurement and the difference between the hot and the ambient. If you plot this, you will see no tendency, just a sort of a random scatter. You could then run a one sample t-test on the differences to see whether there is a significant deviation from zero, but I think there will not be. The differences scatter randomly +/-3 around the zero line so this seems to be practically insignificant.
What you could also do is to measure one sample repeteadly (10 times, say). Those values will scatter as well and with a bit of luck by about the same amount (+/- 3)- Then you might have proven that the differences you see in before and after are simply due to measurement variation and you are done.
Regards and good luck
The relationship does not seem to be that strong.
Spoiler:
Although the correlation is significant, the relationship does not seem strong, in my view.
Yepp, hopefully because there isn't any - it is just measurement noise probably.
Thanks friends... just a question: when a correlation is "strong"?
It is to be expected - it only says that if in HOT the value was higher then it will be higher in AMBIENT as well, same for lower.
Tweet |