Best series to fit a single value ?

Question

Hello, please kindly advice me:

I am trying to find the best fitting mean , measured continuosly in one minute interval to a gold standard (single value). In my example, below: Variable measured at 14:00 - 14:06, which sequence (and they must be in this order) fit best in regards to mean the goldstandard: 41.

I was thinking calculating means from

(14:05-14:06)

(14:04-14:06)

(14:03-14:06)

...

(14:00-14:06)

and compare then all means and lowest Sd to my my gold standard ?

Is there are better or automatic way to do so ?

Many thanks for your advice.

Marc · Answer

Many thanks - please allow me to explain  my problem more in detail.

I have gold standard (paCO2 = 40), which is always 40. In Group 1 - 3 I have one minute measured variables. Which of the groups have a mean closest to the gold standard (40) and the lowest SD ?

- if I try matched pairs marking all groups and paCO2 I get only 5 matched pairs - only if I mark each group individual to paCO2 I get individual matched pairs

My questions:

- can I standardise that I get automatically all matched pairs like shown in picture 2 ?

- how can I get a ranking comparing each group to the gold standard ? - Group 2 has the closest mean to the Goldstandard, but a larger SD than Group 1. Group 3 seems to be the worst  - what test could I use ?

- In other words, how can I check for the best model (group) to fit my Gold standard.

Many thanks again, Marc

Mark_Bailey · Answer

Alternatively, you could use the Matched Pairs platform if you expect only a difference in the mean over the entire range and you want to test for a non-zero difference.

Mark_Bailey · Answer

Assuming that you completely fill the Gold Standard column, then you might try:

* Select Analyze > Fit Y by X.
* Select Variable and click Y.
* Select Gold Standard and click X.
* Click OK.
* Click the red triangle at the top and select Fit Orthogonal > Univariate Variances.

This method is appropriate because you expect a linear relationship and errors in both X and Y variables.