| Author | Message | |||
     David W. Hopkins (Dhopkins) |
All, In writing the article in NIR News Vol. 14 No. 5 pp. 10 � 13, I had some lively discussions with some reviewers on some aspects of the analysis. I hope that we can discuss some of those issues in this discussion thread, and open up to more contributions. In the attached powerpoint, I have included new versions of some of the figures, to clarify points I was making in the article. Slides 1 and 2 show how the second derivative (11-point quartic fit Savitzky-Golay method) is degraded in a 9-point quadratic fit. Toggle between slides 1 and 2 to see this effect. Slides 3 and 4 show the scores plots of Fig. 5, with the colors making the distribution of points much clearer. Slide 5 shows the negative of PC1 and PC2 that were used to make the scores plots. Again, the colors make the loadings clearer, and help make the point that troughs in (-)PC1 and PC2 plots show considerable overlap with bands (negative peaks) in the spectrum of the active, which suggests the reason why samples with highest active content occur in the upper left quadrant of the scores plot, and samples with lowest active content lie in the lower right quadrant. In particular, one reviewer said that the signs of the PCs are arbitrary, so it is not necessary to talk about changing the sign of PC1. However, I maintain that it is important to consider the sign of a PC, when one is considering the position of samples along a PC axis of a particular scores plot. What is your opinion? Finally, the headings of the columns in Table 1 were not proper, and Slide 6 shows the actual statistics that are listed in the table. For instructions on including an attachment file, such as a powerpoint or excel spreadsheet, see the menu item in left sidebar, Documentation/Formatting. I think this will be very useful. Remember, you can still down-load the Shoot-out data from the IDRC website, IDRC-Chambersburg.org and try your own analysis of the data. I think it is a fine data set, and worthy of further discussion. If you would like a copy of the scans of the actives please contact Gary Ritchie, [email protected]. Also, I think it would be okay with him, I would be happy to share the files he sent me. Best wishes, Dave Hopkins [email protected]
| |||
| David W. Hopkins (Dhopkins) |
I did not get the website entry just right. www.idrc-chambersburg.org Hope this works. | |||
| David W. Hopkins (Dhopkins) |
All, It appears that you may not receive the attachment. Please go to the www.nirpublications.com website, select Discussion Group, and under Bruce Campbell's List, find the Calibration Transfer, shootout Discussion area. There you will find the button for the download. If you can take the time while you are there, read the Documentation/Formatting information on the left sidebar to learn how to attach files to your messages. I think this is a really nice new feature for our discussions. Best regards, Dave |
Shootout-DH6slides.ppt