JMP JMP 77 and and Minitab Minitab 15 15 Thomas A. Little Ph.D. 07/07/07 1 TLC, SS0 070402 Audience Description This presentation is designed for those individuals who are interested in understanding the differences in the design, function and capabilities of JMP 7 versus Minitab 15. Particular attention is made to those features and functions used for Six Sigma/Lean project application. Software JMP 7 and Minitab 15. Limitations 762 North 470 East American Fork, UT 84003 1-925-285-1847 [email protected] www.dr-tom.com This presentation is limited to those features and functions of greatest interest to users in the scientific, business, engineering and six sigma/lean communities. An attempt was made to review the features and functions in both applications from a users perspective. TLC actively consults with both applications and finds features and functions in both applications that are best in class. Any disagreements about observations found in this presentation should be addressed to the author who welcomes opposing points of view. 2 TLC, SS0 070402 Presentation Outline Section I General Interface and Ease of Use Section II Lean Six Sigma Activities Define Measure Analyze Improve Control Section III Extended Capabilities Section IV New Features and Conclusions 3 TLC, SS0 070402 JMP Version 7.0 Overview Power JMP provides more analytical tools, graphs, depth, scripting and features that are used to solve real world problems Static and dynamic visualization of data via meaningful graphs and
options. Version 7 added significantly to this capability. JMP is particularly good at large data sets and multivariate modeling JMP benefits from SASs core capabilities and years of development JMP version 7 improves linkage and data transfer to SAS Speed Single define, multiple output All graphs and reports in the same window, powerful table commands not available in excel Control, command function to manipulate them all Ease of Use JMP organization simplifies the windows, text and graphs integrated Simplified interface to complex activities such as Fit Y by X and Fit Model Ease of data and table manipulation. 4 TLC, SS0 070402 Minitab Version 15 Both Minitab (MT) and JMP are far superior for data analysis than using Excel MT is a mature, full featured product with years of user input and product features MT was selected by GE and Honeywell as the early six sigma engine of choice when JMP was just developing version 4. At the time they were correct, MT was the better, more mature product. The world has spun since that time and JMP has surpassed MTs capabilities in all three of the areas of greatest interest to users; speed, power and ease of use. MT release 15 remains a blessing and a curse. Blessing due to its years of application development and familiar tools. Curse due to its old, awkward interface and software design. MT continues to be a much slower application once the data sets rises above 100,000 observations. 5 TLC, SS0 070402 Section I General Interface and Ease of Use General design Windows Organization 6 TLC, SS0 070402 General Design, Tables Minitab uses projects and worksheets as major file formats; where projects are collections of worksheets. JMP has similar capabilities. Table commands for Minitab and JMP are very similar and JMP has some additional table features not found in MT. More table manipulation tools in JMP and more readable file formats. Advantage JMP 7 TLC, SS0 070402 Data Table Size Opening and Manipulating Large Data Sets* File Size (rows)
Time to File Open Time to Display One Histogram JMP Minitab JMP Minitab 1M <1 sec. 13 sec. 1 sec. 90 sec. 5M 5 sec. 15 sec. 6 sec. 100 sec. 20 M 24 sec. Failed. 35 sec. Failed to display Minitab failed to load 20M rows, all 3 columns, only one column loaded. Advantage JMP JMP takes seconds and Minitab takes minutes to manipulate data. If datasets are large as they are in many transactional environments MT is not a tenable solution. Even with moderately sized data tables MT feels slow on response times. *MT JMP evaluation PC used was running Vista, 1.80 GHz Duo, 2GB RAM 8 TLC, SS0 070402 Data Tables and Graphs Linked In MT there is row identification capability; however, no real connection between the graph and table. JMP makes the connection which allows for ease of row location, data and graph manipulation. Major Advantage JMP 9 TLC, SS0 070402 Menus MT displays the analysis method by name. JMP layers the analysis based on one variable, two, paried and multiple Xs and multiple Ys. Menu Pros and Cons Analysis of One
Two Paired or Many variables of any data type. Minitab is easier to use if you are looking for a specific type of analysis by name. JMPs Analyze tools are organized based on single, two, paired and multiple factors. JMP is generalized and easier to learn and remember. This is particularly true of Green Belt level training. Major Advantage JMP 10 TLC, SS0 070402 Graphs and Analysis File: Clean. Minitab uses a separate graph and session window for most of the output. This feature is very annoying in Minitab and slows down the user and the time to analysis understanding. It is a very old school design. JMP keeps all reports and graphs together in one place. Advantage JMP 11 TLC, SS0 070402 Subsets JMP is visual and intuitive when creating subsets. MT does it with formulas, row numbers or brushing. Advantage JMP 12 TLC, SS0 070402 Formulas and Functions JMP has a complete and rich set of integrated functions for data and string manipulation. MT has fewer overall functions and they are spread out and segmented in the Calc function. Advantage JMP 13 TLC, SS0 070402 Section II Six Sigma Activities Define Link to process flow analysis Measure Process capability and MSA Analyze Hypothesis testing and performance modeling Improve
and Design of Experiments Robust Tolerance Design Control SPC 14 TLC, SS0 070402 Define, Process Flow Analysis Minitab and JMP are developing partnerships for linking process mapping, value stream mapping and Lean manufacturing analysis tools into their respective analytical engines. iGrafx for example has both JMP and MT connections. Advantage - Draw 15 TLC, SS0 070402 Process Capability, Minitab Normal Process Capability Sixpack of Cn I ndividual Value I Chart 1 180 1 11 Capability Histogram 1 11 1 1 1 1 _ X=170.62 170 160 UCL=177.95 1 1 1 11 20
40 60 80 1 100 LCL=163.30 1 120 140 160 180 162 165 Moving Range Chart Moving Range 1 1 10 168 177 180 1 1 UCL=9.00 __ MR=2.76 0 LCL=0 1 20 40 60 80 100 120 140 160 180 160
Last 25 Observations 170 180 Capability Plot Within StDev 2.44247 Cp 1.09 Cpk 1.01 CCpk 1.09 180 Values 174 Normal Prob Plot AD: 0.666, P: 0.081 1 5 175 170 Within Overall Overall StDev 3.99757 Pp 0.67 Ppk 0.62 Cpm * Specs 175 File: Cn 171 180 185 Observation 190 195 MTs process potential study is poorly named in this graph. Missing PPM and sigma quality. 16 TLC, SS0 070402 Process Capability, JMP Normal Control Chart Individual Measurement of Cn 1 1 11
180 1 1 11 1 1 175 170 3 .99 UCL=177.95 .95 .90 Avg=170.62 165 1 1 160 .50 11 1 1 15 30 45 60 75 90 120 150 .75 LCL=163.30 .25 .10 .05 180 210 .01 Sample Moving Range of Cn LSL Target USL 2 1 0 Normal Quantile Plot Cn 185 Cn Distributions * 20 *
* UCL=9.00 Avg=2.76 0 Mean LSL 160 Target 170 +3s USL 180 10 160 165 170 175 180 LCL=0.00 15 30 45 60 75 90 120 150 180 210 Normal(170.624,3.99247) Paramete Type P Location Dispersion s Capability CP CPK CPM CPL CPU Index 0.668 0.616 0.660 0.720 0.616 Lower CI Upper CI 0.602 0.734 0.539 0.692 0.596 0.724 0.635 0.805 0.539 0.692 Portion
Below LSL Above USL Total Outside Percent 1.5380 3.2346 4.7726 Benchmark Z Z Bench Z LSL Z USL Index 1.667 2.160 1.847 PPM 15380.212 32345.562 47725.774 Sigma Quality 3.660 3.347 3.167 Control Chart, Sigma = 2.44165 Sample JMPs six graph analysis is hard to find without training; however, it is very good and is easy to interact with. It is a feature under control charts. JMP includes sigma quality in its report and has more secondary options. It allows for nonnormal fit selection on the fly. Advantage JMP % Actual 0.5076 2.5381 3.0457 Overall, Sigma = 3.99247 -3s -3 40 * * Portion Below LSL Above USL Total Outside -2 Count Moving Range of Cn 10 * Fitted Norm Value
162 178 170 -1 30 * Capability Analysis Specification Lower Spec Limit Upper Spec Limit Spec Target -3s LSL 160 Mean Target 170 +3s USL 180 JMPs second capability graph is poorly named. It should be called process potential. Capability CP CPK CPM CPL CPU Index 1.092 1.007 1.058 1.177 1.007 Lower CI Upper CI 0.984 1.200 0.897 1.117 0.957 1.160 1.052 1.303 0.897 1.117 Portion Below LSL Above USL Total Outside Percent 0.0206 0.1261 0.1467 Benchmark Z Z Bench Z LSL Z USL
Index 2.975 3.532 3.021 PPM 206.0636 1260.6901 1466.7537 Sigma Quality 5.032 4.521 4.475 17 TLC, SS0 070402 Nonnormal Capability Fitting Distributions File: Skewed Particles 60 40 20 10 Count USL 20 Gamma(3.79285,1.71318,0) Moments 17.910 17.397 14.691 11.322 8.245 5.959 3.928 2.766 1.757 1.378 1.209 Mean Std Dev Std Err Mean upper 95% Mean lower 95% Mean N Sum Wgt Sum Variance Skewness Kurtosis CV N Missing Fitted Gamma 6.4978344 3.353378 0.1505711 6.7936717 6.2019971
496 496 3222.9258 11.245144 0.8678386 0.4573768 51.607625 0 JMP and MT have similar fitting capabilities, JMP has an interactive interface and an overall better report. Advantage JMP Parameter Estimates Type Shape Scale Threshold Parameter a s ? Estimate Lower 95% Upper 95% 3.7928509 3.3585488 4.2645893 1.7131795 1.5120833 1.9521407 0 . . Note: Unable to converge on all confidence limits. Quantile Plot 11 9 Gamma Quantile Quantiles 100.0% maximum 99.5% 97.5% 90.0% 75.0% quartile 50.0% median 25.0% quartile 10.0% 2.5% 0.5% 0.0% minimum 7 5 3 1 0 0 5 10 15 20 Particles
Capability Analysis Specification Lower Spec Limit Upper Spec Limit Spec Target Value . 20 . Percent %Below LSL %Above USL Actual . 0.000 Overall, Sigma = 3.33646 -3s Mean +3s USL 0 10 20 Capability CP CPK CPM CPL CPU Portion Below LSL Above USL Total Outside Index . 0.928 . . 0.928 Percent PPM Sigma Quality . . . 0.2236 2235.6188 4.343 0.2236 2235.6188 4.343 18 TLC, SS0 070402 Nonnormal Capability in Minitab Process Capability of Particles_ 1 Calculations Based on Gamma Distribution Model USL Process Data LSL * Target
* USL 20.00000 Sample Mean 6.50403 Sample N 496 Shape 3.74418 Scale 1.73710 Overall Capability Pp * PPL * PPU 0.92 Ppk 0.92 Exp. Overall Performance PPM < LSL * PPM > USL 2373.00 PPM Total 2373.00 Observed Performance PPM < LSL * PPM > USL 0 PPM Total 0 0.0 3.6 7.2 10.8 14.4 18.0 MT is missing the sigma quality level and the quantile plot to look at the quality of the fit. The sixpack report is a better option in general when using MT. 19 TLC, SS0 070402 Minitab Pareto Pareto Chart of Causes 300 100 250 Count MT does not allow for easy selection of comparison groups and does not allow for DPU summary tables from the Pareto platform. Cannot directly generate a cost or severity weighted Pareto plot.
60 150 100 40 50 20 0 n s n n ct ct io tio fe ou fe tio s a e e e a n ro D D in l liz e l la n or m a d e o C t a i e lic nt isc Ox M Si M Co Count 110 86 18 17 16 11 Percent 41.0 32.1 6.7 6.3 6.0 4.1
Cum % 41.0 73.1 79.9 86.2 92.2 96.3 Causes g in p Do Percent 80 200 0 10 3.7 100.0 20 TLC, SS0 070402 JMP Pareto Plots 100 250 90 Count 70 60 150 50 100 40 50 20 30 Cum Percent 80 200 03/01/1991 03/02/1991 03/03/1991 03/04/1991 100 90
80 70 60 50 40 30 20 10 0 30 Cum Percent 20 Count Process A 25 15 10 5 0 35 30 100 90 80 70 60 50 40 30 20 10 0 Cum Percent 20 Count Process B 25 15 10 Causes Causes Causes Causes Doping Silicon Defect Corrosion Metallization Oxide Defect Miscellaneous
Contamination Doping Silicon Defect Corrosion Metallization Oxide Defect Miscellaneous Contamination Doping Silicon Defect Corrosion Metallization Oxide Defect Miscellaneous Contamination Doping Silicon Defect Corrosion Metallization Oxide Defect Miscellaneous Contamination Doping Silicon Defect Corrosion Metallization Oxide Defect Miscellaneous Contamination 5 Causes Doping Metallization Corrosion 0 Causes
03/05/1991 35 0 Silicon Defect Oxide Defect Miscellaneous Plots Contamination 10 0 Per Unit Rates Sample Size = 26488 Cause Contamination Oxide Defect Miscellaneous Silicon Defect Corrosion Metallization Doping Pooled Total Count 110 86 18 17 16 11 10 268 DPU 0.0042 0.0032 0.0007 0.0006 0.0006 0.0004 0.0004 0.0014 Lower 95% 0.0034 0.0026 0.0004 0.0004 0.0003 0.0002 0.0002 0.0013 Upper 95% 0.0050 0.0040 0.0011 0.0010 0.0010 0.0007 0.0007 0.0016 JMP allows for easy grouping variables, DPU summary tables and cost and severity weighted Pareto generation. Advantage JMP
21 TLC, SS0 070402 Surface Plots, MT Contour Plot of Yield vs tpd, vph 0.00000010 Yield < 0.0 - 0.1 - 0.2 - 0.3 - 0.4 - 0.5 - 0.6 - 0.7 - 0.8 - 0.9 - 1.0 > 1.0 0.00000009 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 0.00000008 tpd 0.00000007 0.00000006 0.00000005 0.00000004 0.00000003 0.00000002 0.00000001 6 7 8 9 10 11 Surface Plot of Yield vs tpd, vph vph Both MT and JMP have nice surface characterization capabilities. MT is slow to generate and difficult to manipulate. Control over the image is slower and has less options.
1.0 Yield 0.5 0.0 0.00000000 10.5 9.0 vph 0.00000005 7.5 6.0 tpd 0.00000010 22 TLC, SS0 070402 Surface Plots, JMP 11.0 10.0 vph 9.0 8.0 7.0 6.0 1e-8 2e-8 3e-8 4e-8 5e-8 6e-8 7e-8 8e-8 9e-8 tpd 3D visualization in JMP is excellent in either the contour or surface plots. JMP allows for up to 100 gradients and MT allows for only 11 in the contour plot. JMPs Surface Profiler is based on Open GL a full 3D graphics engine. Advantage JMP 23 TLC, SS0 070402 GR&R in MT File: Gage study Gage R&R (ANOVA) for Measurement Reported by: Tolerance: Misc: Gage name: Date of study: Components of Variation Measurement by Part Percent 80 % Contribution
1.5 % Study Var % Process % Tolerance 40 1.0 0.5 0 Gage R&R Repeat Reprod 1 Part-to-Part 2 3 Sample Range Cindy George 9 10 0.5 Cindy George George Operator Tom Operator * Part I nteraction Tom UCL=0.9265 __ X=0.8106 LCL=0.6946 Average Cindy Sample Mean 8 1.0 Xbar Chart by Operator 0.50 7 1.5 UCL=0.292
_ R=0.113 LCL=0 0.0 0.75 6 Measurement by Operator Tom 0.5 1.00 5 Part R Chart by Operator 1.0 4 1.00 Operator 0.75 George Tom Cindy 0.50 1 2 3 4 5 6 Part 7 8 9 10 ANOVA analysis is similar, JMP has the variability graph which is better at displaying variation patterns. MT removes some of the misleading AIAG reports and provides an easier to read report format. MT is missing the secondary breakdown of variation. 24 TLC, SS0 070402 JMP GR&R Functionality JMP has the variability chart that is better for showing variation patterns in the data; however, it is missing the control chart for outlier detection and the summary graphs. JMP needs to add the control chart, summary graphs and secondary breakdown of the
variation patterns to be best in class. Gage R&R Measurement Repeatability Operator*Part Reproducibility Gage R&R Part Variation Total Variation 5.15 63.4896 0.82177 1 2 0.33546 0.35 Variation 0.5751474 0.2918240 0.3454725 0.6709290 0.8164450 1.0567536 % of Tolerance 28.76 14.59 17.27 33.55 40.82 52.84 % Process 31.91 16.19 19.17 37.22 45.30 58.63 which is k*sqrt of V(Within) V(Operator*Part) V(Operator)+V(Operator*Part) V(Within)+V(Operator)+V(Operator*Part) V(Part) V(Within)+V(Operator)+V(Operator*Part)+V(Part) k % Gage R&R = 100*(RR/TV) Precision to Part Variation = RR/PV Number of Distinct Categories = 1.41(PV/RR) Tolerance = USL-LSL Precision/Tolerance Ratio = RR/(USL-LSL) Historical Sigma Variance Components for Gage R&R Component Gage R&R Repeatability Reproducibility Part-to-Part Var Component 0.01697222 0.01247222 0.00450000 0.02513272 % of Total 40.31 29.62
10.69 59.69 20 40 60 80 25 TLC, SS0 070402 Bias and Linearity, MT Gage Linearity and Bias Study for Measurement Reported by: Tolerance: Misc: Gage name: Date of study: 1.00 95% CI Data Avg Bias 0.75 S Linearity 0.131509 0.064619 Reference A verage 0.5 0.55 0.8 0.95 1 1.05 0.25 0 P 0.042 0.017 R-Sq 6.3% % Linearity 18.5 Gage Bias Bias % Bias -0.019444 5.6 -0.027778 7.9 0.111111 31.7 -0.018056 5.2 -0.044444 12.7 0.011111 3.2 -0.086111 24.6 P 0.090
0.339 0.291 0.226 0.144 0.516 0.003 Percent of Process Variation -0.25 20 -0.50 0.5 0.6 0.7 0.8 0.9 Reference Value 1.0 Percent Bias 0.50 0.00 Gage Linearity Coef SE Coef 0.13379 0.06474 -0.18463 0.07619 Predictor Constant Slope Regression 10 0 Linearity Bias The linearity graph in MT is in error. The reference line should be relative to the mean and not to zero. MT does not have the secondary breakdown of bias by part and by comparison group. MT does have the p-values for all of the comparisons which is very desirable. 26 TLC, SS0 070402 Bias and Linearity, JMP Bias Report for Operator Bias Report for Part Bias/Accuracy
0.8 0.6 0.4 0.2 0.0 -0.2 -0.4 Avg Bias 0.02500 -0.01833 -0.06500 1.0 0.8 Bias/Accuracy Operator Cindy George Tom 1.0 0.6 0.4 0.2 0.0 -0.2 Cindy George Tom -0.4 Operator 1 2 3 4 5 6 7 8 9 10 Part 1 2 3 4 5 6 7 8 9 10 Avg Bias 0.11111 0.01111 0.01111 0.02778 -0.02778 -0.08889 -0.04444 0.00000 -0.08333 -0.11111 Part Linearity Study 1.0 Bias/Accuracy
0.8 0.6 0.4 0.2 Standard Value Avg Response 0.50000 0.47222 0.55000 0.66111 0.80000 0.78194 0.95000 0.90556 1.00000 1.01111 1.05000 0.96389 0.0 Linearity -0.2 % Linearity Avg Bias/Accuracy -0.4 .5 .6 .7 .8 .9 1.0 1.1 % Accuracy Process Variation Standard Value t Ratio Measurement = 0.1337949 - 0.1846257 Standard Value Prob>|t| R-Squared -0.065 18.463 -0.00903 2.579 0.350 -2.423 0.017 0.082 Avg Bias Lower CL -0.02778 -0.43030 0.11111 -0.37941 -0.01806 -0.22487 -0.04444 -0.32794 0.01111 -0.38733 -0.08611 -0.45229 Upper CL 0.513260 0.443913 0.197055 0.244745 0.285671 0.332170 JMPs reports are correct and more detailed in general. JMP is missing the p-values for the bias errors. JMP displays the impact to the standard deviation based on rotation effects. Advantage JMP
Which equals Slope * Process Variation 100 * abs(Slope) Bias averaged over all parts 100 * AvgBias / Process Variation Entered on dialog tests H0: the slope equals 0 small pvalues = slope is not likely 0 27 TLC, SS0 070402 Attribute GR&R, MT Date of study: Reported by: Name of product: Misc: Assessment Agreement Within Appraisers Appraiser vs Standard 100 95.0% CI Percent 90 90 80 80 Percent Percent 100 70 60 70 60 50 Ernesto 95.0% CI Percent 50 J uan Appraiser Maria Ernesto J uan Appraiser Maria MT has a very good and very detailed agreement analysis report; however, it is poor on
graphing and labeling of effectiveness. Agreement/effectiveness by part, prob(miss), prob(false alarm), bias report and escape rate are all missing in MT. 28 TLC, SS0 070402 % Agreement Attribute GR&R, JMP 100 80 60 40 20 Juan Maria Ernesto Rater Agreement between & within raters Effectiveness (Agreement to Standard) Agreement Report Attribute Gage Rater Juan Maria Ernesto Gage Attribute Chart % Agreement 100 80 95% Upper CI 82.4890 88.2076 86.6248 % Agreement 53.333 95% Lower CI 36.142 95% Upper CI 69.768 Agreement within Raters 20 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 Part No. % Agreement 95% Lower CI 51.0005 59.0717 56.7146 Number Inspected Number Matched 30
16 60 40 0 % Agreement 68.8889 76.6667 74.4444 Rater Juan Maria Ernesto Number Inspected Number Matched 30 28 30 28 30 29 Rater Score 93.3333 93.3333 96.6667 95% Lower CI 78.6765 78.6765 83.3296 95% Upper CI 98.1523 98.1523 99.4091 Effectiveness Report Agreement Counts 100 80 60 40 20 Juan Maria Ernesto Rater Agreement between & within raters Effectiveness (Agreement to Standard) Agreement Report Rater Juan Maria Ernesto % Agreement 68.8889 76.6667 74.4444 95% Lower CI 51.0005 59.0717 56.7146
Number Inspected Number Matched 30 16 Agreement within Raters Rater Juan Maria Ernesto JMPs attribute GR&R report is very good and covers agreement and effectiveness very well. It is missing bias and escape rate. JMPs graphs are better at showing agreement (blue line) and effectiveness (red line). 95% Upper CI 82.4890 88.2076 86.6248 % Agreement 53.333 Number Inspected Number Matched 30 28 30 28 30 29 Effectiveness Report 95% Lower CI 36.142 Rater Score 93.3333 93.3333 96.6667 95% Upper CI 69.768 95% Lower CI 78.6765 78.6765 83.3296 Correct(0) Correct(1) Advantage JMP Total Correct Correct(0) 25 38 34 Correct(1) 41 50 42
Total Correct 66 88 76 Incorrect(0) Incorrect(1) Grand Total 14 10 90 1 1 90 5 9 90 Effectiveness Rater Juan Maria Ernesto Effectiveness 95% Lower CI 73.3333 63.3802 97.7778 92.2555 84.4444 75.5672 95% Upper CI 81.3762 99.3885 90.5017 Error rate 0.2667 0.0222 0.1556 Misclassifications Standard Level 0 1 Other 0 . 20 0 1 20 . 0 Conformance Report Rater Juan Maria Ernesto P(False Alarms) 0.1961 0.0196 0.1765 P(Misses) 0.3590 0.0256 0.1282 Assumptions
NonConform = 0 Conform = 1 95% Upper CI 98.1523 98.1523 99.4091 Agreement Counts Rater Rater Juan Maria Ernesto Incorrect(0) Incorrect(1) Grand Total 29 TLC, SS0 070402 Context Sensitive Fit Y by X This is where JMP shines over Minitab and provides the user with the proper analysis depending on the data type. JMP automatically switches between four different analytical platforms depending on the column attributes. Advantage JMP 30 TLC, SS0 070402 Correlation Fit Y by X Correlation studies, exploratory data analysis, fit special, group by, etc., this is where JMP outperforms MT on option after option. Advantage JMP File: Factory RSM 31 TLC, SS0 070402 Fit Y by X Contingency Tables Contingency Analysis of Causes By Process Mosaic Plot Causes 1.00 Silicon Defect 0.75 Oxide Defect 0.50 Miscellaneous Metallization Doping Corrosion 0.25 JMP and MT have similar summary table capabilities; however,
MT is missing the visualization graphs. Contamination 0.00 Process A Process B Advantage JMP Process Freq: Failure Count Process Contingency Table Count Contamination Corrosion Doping Total % Process A 86 8 32.09 2.99 Process B 24 8 8.96 2.99 110 16 41.04 5.97 Causes Metallization 5 1.87 5 1.87 10 3.73 5 1.87 6 2.24 11 4.10 Miscellaneous Oxide Defect 8 2.99 10 3.73 18 6.72 42 15.67 44 16.42 86 32.09 Silicon Defect 8 2.99 9 3.36
17 6.34 162 60.45 106 39.55 268 Tests Source Model Error C. Total N DF 6 256 262 268 Test Likelihood Ratio Pearson -LogLike 12.85597 391.44640 404.30237 ChiSquare 25.712 24.743 File: Failures RSquare (U) 0.0318 Prob>ChiSq 0.0003* 0.0004* 32 TLC, SS0 070402 Multiple Regression, N-Way, ANCOVA MT requires detailed statistical and modeling training to remember the names of all of the types of ANOVA. Once the analysis is preformed there is not an easy to use suite of tools and secondary graphs for the user to interact with for further visualization, characterization and optimization. Tools are segmented and not well integrated for optimization. File: cement 33 TLC, SS0 070402 Multiple Regression, N-Way, ANCOVA Prediction Profiler Strength 25.53118 1.762336 35
Simple model definition no matter the data type. 30 25 0.50 Consolidated Brand reinforced Additive 51.01 Humidity .75 1.00 .50 .25 70 .00 65 60 55 50 45 40 standard reinforced Graystone EZ Mix Consolidated 0.00 Desirability 0.462776 1.00 20 Desirability Response Strength 20 20 25 30
35 25 20 24 Strength Predicted P<.0001 RSq=0.82 RMSE=1.7691 25 26 27 28 29 30 0.815622 0.732652 1.769063 25.99761 30 Least Sq Mean 24.510344 25.187449 28.535844 30 25 20 25.5 26.5 27.5 28.5 30 25 20 45 50 55 60 65 70 Humidity Leverage, P=0.0025 Least Squares Means Table Std Error 0.58174366 0.68951528 0.60951464 Mean 24.2011 25.8237 27.9681 Level
Least Sq Mean reinforced 27.313739 standard 24.842019 Std Error 0.58483741 0.50664460 Mean 27.9040 24.0912 Analysis of Variance Source Model Error C. Total DF 9 20 29 Sum of Squares Mean Square 276.88262 30.7647 62.59167 3.1296 339.47429 F Ratio 9.8303 Prob > F <.0001* Parameter Estimates Term Intercept Brand[Consolidated] Brand[EZ Mix] Additive[reinforced] Humidity Brand[Consolidated]*Additive[reinforced] Brand[EZ Mix]*Additive[reinforced] Brand[Consolidated]*(Humidity-51.01) Brand[EZ Mix]*(Humidity-51.01) Additive[reinforced]*(Humidity-51.01) File: cement Estimate 39.078555 -1.567535 -0.89043 1.2358601 -0.254865 -0.21502 -0.590918 0.0187235 -0.105633 0.0848815 Std Error 3.670883 0.473612 0.502441 0.375834 0.073724 0.513278 0.551437 0.094179
0.086185 0.07312 t Ratio 10.65 -3.31 -1.77 3.29 -3.46 -0.42 -1.07 0.20 -1.23 1.16 Effect Tests Source Brand Nparm 2 DF 2 Sum of Squares 84.839165 F Ratio 13.5544 Prob > F 0.0002* Prob>|t| <.0001* 0.0035* 0.0916 0.0037* 0.0025* 0.6797 0.2967 0.8444 0.2346 0.2594 Leverage Plot 35 40 Additive Leverage, P=0.0037 Least Squares Means Table Level Consolidated EZ Mix Graystone Brand*Humidit Leverage Plot 35 24.5 Brand Leverage, P=0.0002 Summary of Fit RSquare RSquare Adj
Root Mean Square Error Mean of Response Observations (or Sum Wgts) 30 Brand*Additive In addition to the detailed statistical summary tables JMP offers a full suite of graphs for visualization, characterization and optimization. Advantage JMP 35 Strength Leverage Residuals 25 Leverage Plot 35 Strength Leverage Residuals 30 Humidity Leverage Plot 35 Strength Leverage Residuals 35 Strength Actual Additive Leverage Plot Strength Leverage Residuals Brand Actual by Predicted Plot Strength Leverage Residuals Whole Model 30 25 20 23 24 25 26 27 28 29 30 31 30 25 20 25.0 Brand*Additive Leverage, P=0.3287 Least Squares Means Table Level Least Sq Mean Consolidated,reinforced 25.531184 Consolidated,standard
23.489504 EZ Mix,reinforced 25.832392 EZ Mix,standard 24.542507 Graystone,reinforced 30.577642 Graystone,standard 26.494047 Std Error 0.8448547 0.8660803 1.1401751 0.7918990 0.8031139 0.9428028 34 TLC, SS0 070402 25 Design of Experiments - Design DOE in Minitab is awkward to use for designing experiments as it does not allow for the direct design of the experiment in line with the problem that needs characterization. Minitab uses a candidate points method for customization and augmentation. This is very old school and tedious for the user. Covariates are not part of the design, they are secondary in the analysis. Minitab does not allow for correct factor identification when designing the experiment. There are many more factor types than those allowed by MT. MT fails the ease of use test for DOE. File: Yield 35 TLC, SS0 070402 DOE Analysis, MT Analysis flow MTs analysis tools for DOE are segmented, do not flow well and the optimizer is missing a more intuitive set of controls for constraining, fixing, optimizing and predicting the response. MTs DOE design and analysis flow is segmented, complicated, not seamlessly integrated and has too many steps. 36 TLC, SS0 070402 Design of Experiments in JMP JMP custom designs match the problem. Any combination of factors, factor types, covariates, blocking sizes, categorical factors and mixtures with a minimum sample size. Simple to define the model terms to be characterized. Allows the most flexible environment for DOE treating the engineer and scientist as the customer. JMP is best is class for DOE.
JMP wins on DOE ease of use. In JMP the DOE design always fits the problem. 37 TLC, SS0 070402 DOE Analysis in JMP is the Same Fit Model Engine Output 1275 20.74915 Prediction Profiler 1600 1200 Cracks 4.8 1.037457 Diameter 2.499 0.025936 800 2.65 2.55 2.45 2.35 20 15 10 5 275 Temp 7.5 Time .25 .50 .75 30 .00 25 20 22.5 Pressure 1.00 150 Speed 200 250 260 270 280 290 300 5 6 7 8 9 10
15 175 150 125 100 0.00 0.50 1.00 Desirability 0.053217 0 Desirability In JMP learn one set of tools and use them for a variety of characterization, DOE, modeling problem solving activities. JMPs profiler allows for improved visualization and control of the transfer functions. Major Advantage JMP 38 TLC, SS0 070402 JMPs Simulator Linked to Transfer Functions Optimize performance, improve robustness and predict full distribution at target. MT does not have this capability. Set and evaluate tolerances. Major Advantage JMP Diameter Capability Analysis USL 75000 Count LSL 50000 25000 Specification Lower Spec Limit Upper Spec Limit Spec Target Value 2.51 2.57 . Portion Below LSL Above USL Total Outside % Actual 0.4050 0.7275 1.1325 Overall, Sigma = 0.01178 Capability CP
CPK CPM CPL CPU 2.48 2.5 2.52 2.54 2.56 2.58 2.6 -3s LSL 2.48 2.52 Mean +3s USL 2.56 2.6 Index Lower CI Upper CI 0.849 0.000 0.850 0.846 0.844 0.847 . . . 0.853 0.851 0.854 0.846 0.844 0.847 Portion Below LSL Above USL Total Outside Percent PPM Sigma Quality 0.5271 5270.6210 4.058 0.5586 5586.2125 4.037 1.0857 10856.834 3.795 Benchmark Z Z Bench Z LSL Z USL Index 2.295 2.558 2.537 39 TLC, SS0 070402 Power and Sample Size JMP has sample size calculation for counts per unit and for estimating the standard deviation. MT identifies sample size for replicates for two specific forms of DOE and JMP does not. JMP also has a sigma quality converter and calculator.
Minor Advantage JMP 40 TLC, SS0 070402 SPC JMP 6 to Minitab 14 Comparison 11/22/2005 SPC Control Charts Control Charts for Subgroups Xbar R Xbar S Presummarize Delta to Target, subgroup Z subgroup Control Charts for Individuals Run Chart I/MR Z/MR individual Delta to Target, individual Levey Jennings Control Charts for Small Mean Shifts UWMA (moving average) EWMA CUSUM Control Charts for Attributes P NP C U Multivariable Control Charts T2 Multivariate EWMA JMP 6.0 MT 14.1 Y Y Y N N Y Y Y N Y Y Y N N Y Y Y Y N N Y Y Y Y Y Y Y
Y Y Y Y Y Y Y N N Y Y 11/22/05 MT and JMPs capabilities are quite similar. MT offers more charts; however, JMPs charts are easier to manipulate and are better for larger data sets. JMP needs to add the short run Z and delta to target charts. Both platforms allow for phased control charts to show before and after effects. Advantage - Draw 41 TLC, SS0 070402 Section III Extended Capabilities Reliability Multivariate Time Series Graphs Advanced Modeling Summary 42 TLC, SS0 070402 Reliability MT offers reliability planning tools for sample size determination and JMP does not. JMP has stronger modeling and multivariate tools for reliability modeling. Advantage - Draw 43 TLC, SS0 070402 Multivariate JMP has a richer set of tools for multivariate analysis. Factor analysis and principle components analysis are in the multivariate platform and are harder to locate from the menu. Advantage JMP 44 TLC, SS0 070402 Time Series JMP and Minitab similar tools and capabilities. JMP has a few more options and the ease of use and graphical
manipulation makes it superior to MT. Minor Advantage JMP 45 TLC, SS0 070402 Graphs JMP offers similar graphs to MT; however, it outperforms in the profiler, contour profiler, surface plot and custom profiler options. MT does not have the same rich tools for optimization and robust design. Advantage JMP 46 TLC, SS0 070402 Advanced Modeling Tools JMP offers a much richer and versatile set of modeling tools and analytical methods. Neural nets, recursive partitions and nonlinear modeling are all available modeling tools in JMP. Advantage JMP 47 TLC, SS0 070402 For A More Detailed Comparison JMP 6 to Minitab 14 Comparison 11/22/2005 Product Features File and Data access Table design and tools Supporting file formats Large data table manipulation (1M rows +) Database connection Project file management Customization Programmability, scripting Menus (names and graphics) Toolbars Keyboard commands Full automation Ease of Use JMP Starter Graph Manipulation Menus Help functions Context sensitive help Toolbars Graph and data table link Documentation Dynamic graphs using scripts Integrated graphs and reports Data editing and modification JMP 6.0 MT 14.1 A A A A no feature B B D B+ A
A A A no feature A B A A A B A A A B A A A B A A A no feature C B A no feature B C A no feature C A For a more detailed comparison of JMP versus MT take a look at the JMP 6 to MT 14 comparison table. 48 TLC, SS0 070402 Summary JMP is in general a superior product JMP is world class for regression, modeling, DOE, and simple studies such as process capability and MSA and the user interface is very well designed JMP is easier to use, more powerful, much faster in completing analysis of data and needs to address some of the minor gaps identified in this comparison Having two great applications is good for the market and keeps both applications improving to meet customer needs and expectations MT is a good application and has a rich set of tools. JMP is a great application and has an overall better designed and better integrated tool set. Helping companies understand why Excel is not enough for analysis is the greatest opportunity Minitab must address the ease of use, some missing tools and speed issues. 49 TLC, SS0 070402 762 North 470 East American Fork, UT 84003 925-285-1847 [email protected] www.dr-tom.com 50 TLC, SS0 070402
SHRM/Achieve Survey Findings: Post-secondary Certificates ...
This is the sixth part of a series of SHRM/Achieve survey findings titled "Changing Employee Skills and Education Requirements."These results look at post-secondary certificates/ credentials ...RNA Synthetic Biology - International Genetically Engineered ...
RNA Synthetic Biology Farren J Isaacs, Daniel J Dwyer, & James J Collins Nature Biotechnology May 2006 iGEM 2010 Journal Club 7/7/2010 RNA RNA RNA RNA Antisense RNAs Riboregulators sRNAs (small regulatory RNAs) miRNAs siRNAs Riboswitches Ribozymes Controlling Gene Expression...