Statistical Methods in Chemistry Benjamin
N. Nelson,
Research and Development Department, The Youngstown Sheet
T
review continues the coverage of statistical developments and applications in chemistry and other sciences for the period October 1962 to October 1963. Some of the newer books are mentioned, but particularly, periodical articles in the vgited States literature which cover applbations other than those of a routine nature. Foreign literature and meeting transactions or documents of limited distribution are not mentioned. HIS
BIBLIOGRAPHIES, OTHER REVIEWS, BOOKS, A N D JOURNALS
The previous review (109) and those before it have listed abstract journals designed to keep one current with the literature. Industrial and Engineering Chemistry, in an annual review (129) oh mathematics in chemical engineering, has updated developments from past reviews. Among more basic bibliographies are books on basic texts and monographs on statistical methods covering the period 1945 to 1960 (42) and on statistical literature covering the period 1950 to 1958 (89). I n periodicals, a 1930-1957 bibliography on frequency functions, moments, and graduation (57) and a supplement on 1958-1960 statistical literature (68) have appeared, a list of bibliographies for statisticians (124) and additions (126)) and a bibliography series in a paper journal on precision, analysis of test data, and applied statistics (22,23). Connors completed a series on statistics in one journal (60, 51) and .4ndersen began a series in another (3-14). Basic books include a guide to tables ( V ) ,special table compilations (64, 120), use and abuse of statistics (122), experimental statistics (108), total quality control (67), operational research (69), and statistics in research (117). I n more specific areas are nonparametric statistics ( f 3 3 ) , applications in chemical analyses (for), and a manual on interlaboratory studies ( I ) . STAT ISTICAL APPLICATIONS, EVALUATIONS, A N D INTERPRETATIONS
Good (72) has discussed some basic ideas in an article entitled “A Classification of Fallacious Arguments and Interpretations” and Olmstead ( 1 12) 344 R
ANALYTICAL CHEMISTRY
& Tube Co., Youngstown 1, Ohio
discussed statistical evaluations dealing with making discoveries and predictions. An article has appeared on evaluating “messy” d a t a (74). Price (121) has dealt with applications and the statistical approach in regard to frequency of use in scientific papers. Kennard (90) has interesting comments on strategy in use of statistics. STATISTICAL A N D QUALITY C O N T R O L APPLICATIONS T O CHEMISTRY, ACTUAL A N D POTENTIAL
Precision and Accuracy and Variance. Eisenhart (63) has dealt with some fundamental ideas on precision and accuracy as they relate to instrument calibration systems. As t o nomenclature, there are still various ways of expressing precision. One article by the ASTM D-2 Committee (103) speaks of repeatability and reproducibility, for example, two terms not as yet universally defined, Andersen has a new series on statistics in chemical engineering, in which one article is on estimates and tests of variance ( 8 ) . Youden deals with precision and error estimations in two articles (141 , 144). Analysis of Variance and Designed Experiments. I n t h e determining of sources of variation by analysis of variance and designed experiments, basic articles have been written by Connors in his series on statistical design, Andersen (13, l & , Scott (126), Bingham (26), nicking and Gillespie ( 2 4 , and Kennard (90). Application of analysis of variance has appeared in articles on regression (47) and analysis of curve families (101). Problems of bulk sampling have been discussed by Bingham (29) and Duncan (60). Split plot designs have been applied (48, fog), and latin square (115). I n interlaboratory or collaborative testing, a number of papers have utilized analysis of variance ( 1 , 95, 96, 97, 130, 142). General application journal articles have appeared on agriculture and food (41, 48, 62, 77, 134, 140), paints (379, leather (104, 113, 114, 115), ceramics (21, 92), paper (95, 96, 97, 130), and oil (87). There have been chemical engineering studies (81, 91) and articles on analytical
chemistry and physical tests (1, 43, 63, 76, 110). The class of response surface designs by Box and coworkers has been dealt with in basic papers (15, 16, 17, 20, 24, 27, 32-36, 39, 40, 44, 63, 64, 71, 78, 79, 83, 90, 93, 111, 123, 127, I S ? , 139, 146), and in papers pertaining to computers and instrumentation (15, 20, 46, 63, 1#), organic reactions (16, 16, 18), inorganic applications (17, 24, go), and a metallurgical experiment (28). In the field of chemical processing and chemical engineering, there have been numerous papers (I6, 17, 18, 10, 27, 31 , 39, 44, 54, 69, 78, 79, 82, 93, 106, 111, 128, 139). A topic relating to response, aiming for conditions of optimum production, is evolutionary operation (EVOP), which has been discussed as an effective aid to systematic process improvement (28, 132). Experience in its use in chemical processing has been the subject of one paper (65). A method of computer plant simulation has been compared with the techniques of EVOP and CHFOPS (80).
Correlation, Regression, Covariance. Andersen has written two basic articles on correlating experimental data in chemical engineering (10, 11). He has dealt with rank correlation (12) and its application in a paper on leather (116). A nomograph has been published to aid in computing partial correlation coefficients (99). There have been correlation applications, such as one relating detergency performance to chemical properties (87) and another on showing up trends in data (100). Basic articles have covered regression (10, 47), multiple regression ( I f ) , and the use of regression in screening experiments (63)) in judging compatibility of data with theory (56), and in analyzing families of curves (101). There have been applications in weighted least square (13’7) and nonlinear curve fitting (146). Hoerl has written of another computational procedure called “ridge analysis” (79), which is applied to regression-type problems. There have been regression applications in chemical processing (49, 5?2, 80, I O @ , agriculture (48), and paper (100). A covariance application has appeared in a spectroscopy paper ( f 10).
Miscellaneous Stiitistical Topics. On t h e subject of distribution theory and probability Deming has published a bibliography (57)and Langton (94)has written on deciding on goodness of fit of a set of data with model distributions. Eisenhart has used d stribution curves in comparing precision and accuracy (63), Youden (144) in a paper on systematic errors, and Wescott (136) in regard to specifications. Andersen (12) has compared distributions by nonparametric statistics and Walsh (133) has a handbook on that subject. On the subject of rejection of outliers, Pierson (120) has published tables, and Youdm (148) and Burns and Muraca (43) have written on comparing laboratories. Gnanadesikan (71) has writte i on applications of multivariate analysis, Ornes (115) and Youden (142) on ranking techniques, and Osinski (116) on using range to compute standard deviations. Nimeroff and others (110) have applied Hotelling’s T ter,t in spectroscopy and Connors (50) in tolerance limits. On the topic of sensitivity, there have been articles by Lashof (96) and Tappi (130). -4 sequential search technique has been the subject of a paper by Archer and others ( 1 L ) . Quality Control. .4 book by Feigenbaum (67) has been mentioned. Bingham (25) has written on general applications in the chemical industry, Kleinkauf (92) in ceramics, Bingham (27) in coated abrasives, and Keeler (88) in spectroscopy. On the subject of control charts, articles have appeared on specificationoriented charts (65),and applications in chemical processing $5, 38, 45, 68, 131), ceramics (92), coated abrasives (27), paper (96), agriculture ( 7 7 ) ) and spectroscopy (88). On the subject of cumulative sum c h u t s , which was introduced in the previous review ( l o g ) , there has been a series of articles of basic interest (45, 66, 68, 84, 85, 86, 118, 11