Choosing Software for Statistical Design of Experiments and Regression Analysis

Instructions:
Print out this form and then use it to evaluate your software needs.

	Yes No
1. Is it menu driven?	______	______

2. Can it be used without learning how to write program instructions?	______	______

3.1 Has a scientist or engineer tried the menu system and said s/he liked it?	______	______

3.2 Has a statistician tried the menu system and said s/he liked it?	______	______

4.1 Does the software provide the following types of experiment designs:
a) screening (Plackett Burman) for up to 43 factors	______	______
b)interaction (cube corners); factorials and fractional factorials for up to 10 factors	______	______
c1) quadratic (cube plus star) for a cubical region of interest for up to 6 factors	______	______
c2) quadratic (uniform shell) for a spherical region of interest for up to 8 factors	______	______
d) axis designs (A,B,C,D points) for screening mixture components	______	______
e) simplex lattice designs for mixtures	______	______
f) mixture designs created from lower and upper limits (vertices and centroids)	______	______
g) designs for both mixture and process variables in one experiment	______	______

5. Does the software provide a scrambled run order, that is, a random run order tested and replaced if it turns out to be a bad one?	______	______

6. Are transformations of the factors and responses available?	______	______

7. Can the data be edited:
a) to indicate missing values	______	______
b) to correct entries	______	______
c) to add or delete rows and/or columns?	______	______

8. Does the software retain and display for your convenience the data and model files most recently selected?	______	______

9. Can your data be:
a) keyed in	______	______
b) imported (as from Lotus)	______	______

10. Can the following models be fitted:
a) main effects only (linear model)	______	______
b) main effects and interactions (cross products)	______	______
c) full quadratic	______	______
and for mixtures:
d) linear	______	______
e) quadratic	______	______
f) special cubic	______	______
g) cubic	______	______
h) quartic	______	______
i) Becker H(1), H(2), H(3)	______	______
j) Inverse and for both process and mixtures variables: k) main effects	______	______
l) quadratic	______	______

11. Can you customize the model by suppressing selected terms?	______	______

12a. Can selected observations be withheld from the regression to use for testing the prediction of the fitted model (or for other reasons)?	______	______

12b. Are confidence limits on the next observation computed to assist in testing the prediction quality?	______	______

13. Can contour plots be made for:
a) process factors	______	______
b) mixtures (triangular)	______	______

14a. Does the software find groups of replicates and compute standard deviation for each group of replicates?	______	______

14b. Is the pooled standard deviation computed?	______	______

15. Is the raw data printed to facilitate proofreading?	______	______

16. Is the data standardized:
a) coding factors to the -1 to +1 range (or similar standardization)	______	______
b) by coding to pseudocomponents when there are limits on mixture components	______	______
c) by standardizing each column of the expanded X list (in mixture cases) before computing b's (to avoid numerical problems)	______	______
17. Is standard deviation:
a) computed for each group of replicates	______	______
b) pooled when there are two or more?	______	______
c) tested for uniformity by Cochran's test?	______	______
d) plotted as log s versus log (Y bar) to provide help with transforming Y to get uniform standard deviation?	______	______
e) displayed with confidence limits?	______	______

18. Are confidence limits for the average at each trial provided?	______	______

19. Are confidence limits on b coefficients provided?	______	______

20. Is the variance inflation factor provided for each b coefficient to show whether or not that b can be estimated safely using the data and model at hand?	______	______

21. Does the software refrain from providing step-wise regression when the data come from a good experiment design?	______	______

22. Does the software show residuals:
a) plotted against run order	______	______
b) plotted against Y predicted	______	______
c) plotted against each factor	______	______

23. Are confidence limits for the long term average response given for every point in the X-list whether data is available at that point or not?	______	______

24. Are confidence limits for the next single Y provided at every point listed in the data to assist with testing model fit?	______	______

25. Are tolerance limits calculated, at least approximately for each point in the data list?	______	______

26. Are wild value tests done on groups of replicates?	______	______

27. Is there an option to scatter plot the responses against each other as well as against the factors?	______	______

28. Will the program search the entire region of interest and select the points for which all responses meet their goals?	______	______

29. Does the user have control of the values of the contour levels on the contour plots?	______	______

30. Can the user get a quick look at the contour plots on screen before taking the time to print them?	______	______

31. Do the contour plots show a perspective view (three- dimensional) of the response surface and are the contour lines shown?	______	______

32. Do the contour plots also show a view from above in two dimensions of the response surface with contour lines drawn and labeled?	______	______

33. Does the software show on the page with the contour plot, what model is being plotted, the b coefficients, and the name of the X, Y list these coefficients come from?	______	______

34. Does the software give confidence limits on the same page as the contour plot so that the uncertainty in the prediction of the long term average performance is displayed?	______	______

35. Does the software provide blocking factors for interaction designs?	______	______

36. Does the software print a worksheet suitable to use to record results when doing the runs?	______	______

If the answer is YES to all these questions, then the software is the right one to plan and analyze statistical experiment designs. For STRATEGY(tm) (The Experiment Strategies Foundation) the answer is yes to every question above. Get a copy of STRATEGY(tm) to get more work done in less time. Select "Order Form for Software, Books and Training."