experimental design

[gcguy]

I am planning to look at design software. The exercise I am planning is to look at the impact of variables on results. The sample I am looking at will have 4 significant contributors to the one component I wish to examine.
i.e I have 4 components, I wish to develop a calibration model for one component and I need to test for interference from the 3 other components that will vary by 10%. My plan is to look at low, medium and high level of each of these, with each component varying against each other including the compound of interest.
What I am looking for is software to help with this and future projects in terms of experimental design and I would appreciate input from anyone who has used software for this.

Thanks
GCguy

[Don_Hilton]

Depending on what it is that you are doing, simple statistics can be done in Excel or a similar spreadsheet program. For more complex analysis there are a range of statisitcal packages out there. For price and power, "R" is a good option - it is free and well respected in the statistical community. The user interface is not easy and you have to know something about the statistics you are trying to do to set up the calculations. (But this last point echos discussions on this forum about people who do not understand the chemistry associated with sample prep and chromatography setting up chromatographic systems and methods.)

For more on R see http://www.r-project.org/

[krickos]

Hi

Been a while since I used chemometrics/experimental designs but i liked "Modde" (http://www.umetrics.com/modde). Used to have it availeble on my PC but not anymore.

No clue how price worth it is but it used to be quite easy to set up reduced or full factor designs. As Don indicated it does require that you know what your doing and is trained to evaluate the significance of the contributuions of the various factors alone and combinations.

[DR]

Minitab, JMP, xstat...

[gcguy]

Thanks for the input. I am looking more for an easy way to set up and plan the dilutions and samples in order to perform the model building. A means of producing tables or spreadsheets which will calculate what solutions to make up to perform the experiment rather than the analysis of data generated from the experiment. Software that will allow nested (?) tables and experimental design would be ideal.

Sorry if this sounds a bit basic but I have limited stat knowledge.

GCguy

[Don_Hilton]

For the experimental design, there is an old classic stat book. "Statistics for Experimenters" by Box, Hunter, and Hunter. It helps with experimenal design and with what the data from the designed experiment mean. I got a used copy of an older edition as a used book for a low price some years back - and in going on line to be sure I had the title correct, I discovered that this book is still in print as a later edition. (It's been out there for some 30 years.)

When you pick the design, you may find that calculation of design parameters will work quite well in a spreadsheet. The later edition does address computer analysis of statistics - but the edition I have uses the old pencil, paper, and a calculator.