Section 5 Slides rev 3.1 CSMO

Section 5 Slides rev 3.1 CSMO Section 5: White Paper Template Change Control for Materials MTL008986 Rev 3.1 (CSMO) May 2002 1 The White Paper Template , Use the white paper template for the forum the change will be presented at o Templates are available on that CCB’s web page , The PCCB uses different WP templates , Materials uses a "combined" white paper template (CSMO/ATMO) which will be discussed in this class In this section, we will examine the white paper template and discuss how to complete it. Occasionally additional foils are presented during the template instruction to further explain specific portions of the template. These pages are indicated with the heading “White Paper Notes” at the top of the page. Pages without this heading are directly from the white paper template. Change Control for Materials MTL008986 Rev 3.1 (CSMO) May 2002 2 MCCB White Paper Horizon Schedule #: Assigned by MHS web tool () Title of the Change: Brief description of the change. PWP Date: Revision: 0 Will this PWP generate multiple FWPs? Yes No FWP Date: Revision: 0 Update revision number each time a submission is made to the CCB. It is your choice whether or not you prefer to revise the submittal date. Supplier Change Classification: MCCB: Class A Class B Class C Class C-N Refer to Spec 08-2091 for definition of Change Classifications Intel PCCB (if applicable): Class I Class II Class III Class III-N Class IV PCCB Horizon Schedule #: Classification Justification: Provide information on potential impact, risk assessment or other data available to justify change classification. Commodity: List all commodities affected Supplier: List all Suppliers affected Intel Sites Affected: List the VF sites that are affected by the change or run product affected by the change Divisions Affected: List the Intel Divisions affected by the change. Products Affected: List the Intel Products affected, Platform, part numbers, etc. Name of Originator: Put name of person who originated this paper and an Intel change owner name if a supplier originated it Reason for Change: (Brief) Some examples are: Capacity Increase, Productivity Increase, Quality Improvement, Yield Improvement, Cost Reduction, EHS, etc. Change Control for Materials MTL008986 Rev 3.1 (CSMO) May 2002 3 Change Description: Explain what is changing, describe before/after differences, consider all aspects of possible change: tooling, equipment make/model, chemistry, location, etc. Enter Change Description diagrams, sketches, or flowcharts here. Delete this page if not needed. Add in tables, diagrams, etc to document and illustrate the change Change Control for Materials MTL008986 Rev 3.1 (CSMO) May 2002 4 Change Team: The first column, “X if req?d” will indicate those functions that are on the change team. The X?s already there indicate the standard set that should appear on every change team. Additional members will be added depending on the change. The “PWP Approval” column will be used to indicate when that person has given PWP approval. The “FWP Approval” column will be used to indicate when that person has given FWP approval. X if Area PWP FWP Name req’d. Approval Approval Intel Sponsor Supplier Sponsor QRE Statistics JQT List Members Below by Site CR CV IR KM CAT/ATD CSMO ATD/MTD AMCL Design Finance Integration VF Module/JET Other ECO /ECR Classification*: BOM Change New part number *A change checked here requires Configuration Management on the Change Team Proposed Implementation Date: WW Change Control for Materials MTL008986 Rev 3.1 (CSMO) May 2002 5 Risk Assessment: Low Medium High Consider all aspects of the change such as availability, technology risk, quality, reliability, customer risk, etc. associated with implementing the change or not implementing the change. Refer to “Materials Risk Assessment” (Spec.# BS-CSG-0004) for a description of each risk level. Explanation of Risk Assessment: Whether you check Low, Medium or High above, explain why it is that level of risk. Implementation/Synergy: This table should contain only those who are responsible for implementation of the change and/or ensuring synergy across multiple factories (Intel and/or supplier). This is *not* the change team. “PWP Synergy” and “FWP Synergy” columns should be used to indicate that person is aware of and agreed to that ownership. Site Sponsor PWP FWP Brief Description of Support Synergy Synergy THESE TWO TABLES SHOULD BE USED ONLY FOR AR?S ASSIGNED BY THE MCCB. THEY SHOULD NOT BE USED TO TRACK AN AR FOR SOMEONE PRIOR TO SUBMISSION OF THE PAPER. PWP Actions required by MCCB Owner Due Completed WW WW WW WW FWP Actions required by MCCB Owner Due Completed WW WW WW Change Control for Materials MTL008986 Rev 3.1 (CSMO) May 2002 6 Use the Concerns/Consideration sections below to identify and discuss any additional details about the change and the qualification strategy. For example, explain: , business implications of implementing versus not implementing relationship of this change to other changes , , rationale for selection of quality characteristics , logic of qualification plan , critical logistical considerations or hinge factors , physical mechanism and models Supplier Concerns/Considerations: Intel Concerns/Considerations: Customer Concerns/Considerations: Other Concerns/Considerations: Inventory Control: Note: Revenue product refers to Intel?s shipments to its customers. Check the appropriate box in the PWP column based on what you plan to do at time of PWP submission. Then, when the FWP is submitted, use the FWP column to indicate what actually happened. Only one box should be checked in each column. PWP FWP Revenue product will not be (was not) used. Revenue product will be (was) used, but will be (was) held until FWP approval. Revenue product will be (was) used and shipped prior to FWP approval. Provide explanation if different: Explain reason for changing strategy and decision process used to support it. Change Control for Materials MTL008986 Rev 3.1 (CSMO) May 2002 7 GANTT CHART (Complete for PWP and update as necessary for FWP) KEY: C = Complete; I = In Progress; N = Not Done; NA = Not Applicable WW WW WW WW WW WW WW WW JET FWP Approval MCCB FWP Approval Spec Update Training Guides Training Implement @ Supplier Site/Prod. state site/ location Site/Prod. Site/Prod. Site/Prod. Implement @ Intel Site/Prod. state site/ location Site/Prod. Site/Prod. Site/Prod. Site/Prod. Site/Prod. Other DCCB Buy-off RFC ECO Approved ECO Effectivity Change Control for Materials MTL008986 Rev 3.1 (CSMO) May 2002 8 White Paper – Notes Change Proliferations: In cases where it is not possible to simultaneously implement a change at all sites: , The proliferation plan must be clearly defined in the PWP. This plan must comprehend equipment and material handling requirements. PWP Modifications: Any significant change to an approved PWP requires re-approval by the Supplier and Intel CCB. This includes: , Change in scope , Change in sites affected , Change in the Experimental Strategy Data Sharing , Data from other sources may be used to justify the change o Technical Reports o Other White Papers o Other products or platforms , Applicability of the data must be justified , Shared data used in the qualification should be described and documented in the White Paper Change Control for Materials MTL008986 Rev 3.1 (CSMO) May 2002 9 KEY INDICATOR SUMMARY (Complete for PWP and update as necessary for FWP) No Change – Decrease -- Increase INDICATOR: SUPPLIER INTEL IMPACT IMPACT Process Capability No Change No Change Yield No Change No Change Output Quality No Change No Change Reliability No Change No Change Productivity No Change No Change Capacity No Change No Change Cost No Change No Change Safety/Ergonomics No Change No Change Return on Investment No Change No Change Availability No Change No Change Change Control for Materials MTL008986 Rev 3.1 (CSMO) May 2002 10 COMPLETE TABLES A THROUGH D FOR PWP DO NOT CHANGE FOR FWP TABLE A PROCESS FACTORS Process Factor Present Value Proposed Value List out all aspects of the process related to the scope of the change, whether they change or not. Process factor: ， Parameters whose values are changing ， Process factor values completely describe the change ， Present value describes current condition ， Proposed value describes new condition ， Examples: procedures, equipment, parameter settings, materials, hardware, software, metrology, lot size, etc. Note: Always indicate the unit of measure for the data indicated. Change Control for Materials MTL008986 Rev 3.1 (CSMO) May 2002 11 COMPLETE TABLES A THROUGH D FOR PWP DO NOT CHANGE FOR FWP TABLE B QUALITY CHARACTERISTICS Quality Measurement Ideal Present Present Expected Characteristic Method Target Mean Std Dev Change Quality characteristic: ， Output response ， Any measure of the product or process that may be affected by the change ， May include: module (process) key manufacturing responses, line yield, product quality, product reliability, manufacturing /operations responses, etc. ， Examples: , reliability test performance (HAST, temp cycle) , yield (mechanical, electrical, visual) , defect level (voids, blisters) , DPM (visual inspection, electrical outgoing) , key process parameter (coplanarity, wire pull, thickness) Measurement method: ， System used to quantify the change in or performance of the quality characteristic ， Provide clarification on defect categories or failure rate of concern ， Define „failure? for pass/fail type data ， Be specific in the type of data to be gathered and compared Ideal target: ， Value of the quality characteristic at its highest level of quality ， May not be achievable but reflects the most desirable direction for that quality characteristic to move Change Control for Materials MTL008986 Rev 3.1 (CSMO) May 2002 12 Present mean and standard deviation: ， Most recent measures of quality characteristic at current process factor setting. If data is not available, experimental strategy might call to collect it. In this case, present values should be TBD. Expected change: ， Indicate direction quality characteristic is expected to move ， Always measured relative to current process except when there is no current process data ， Possible Options: No change (no expected change in the characteristic), Better (characteristic moves toward the ideal target), or Worse (characteristic moves away from the ideal target). Note: Always indicate the unit of measure for the data indicated. Change Control for Materials MTL008986 Rev 3.1 (CSMO) May 2002 13 COMPLETE TABLES A THROUGH D FOR PWP DO NOT CHANGE FOR FWP TABLE C NOISE FACTORS Noise Factor Range of Values /Scope Range Evaluated Noise factor: ， Process or environment variable that naturally changes during manufacturing operation ， May influence the measured value of the quality characteristic ， Controlled or uncontrolled: controlled factor is one whose value may be assigned at any time to any level (e.g. material lot), and uncontrolled factor is one whose value is random & may assume any level in the manufacturing environment (e.g. humidity) ， A quality characteristic at one module can be a noise factor at a downstream module. Range of values /scope: ， Entire set of noise factor values over which the change must perform o For discrete, controlled factors; list the number of values available o For continuous, uncontrolled factors; list the consumption (or replacement) rate OR size Range evaluated: ， Subset of noise factor range to be evaluated in experiment ， Full range of noise factor is not always evaluated in an experiment strategy ， Justify elimination of significant noise factors in Other Concerns /Considerations section of PWP. ， Examples: Change Control for Materials MTL008986 Rev 3.1 (CSMO) May 2002 14 o Number of machines o Environmental control (temperature, humidity) o Product type o Raw material batch o Shifts, technicians o Tools, fixtures, manufacturing lots o Number of VF manufacturing lines Note: Always indicate the unit of measure for the data indicated. Change Control for Materials MTL008986 Rev 3.1 (CSMO) May 2002 15 White Paper – Notes Experimental Strategy - Qualification Plan Objective of Experimental Strategy: , Define Quality Characteristic performance for the proposed change in Process Factors across all Noise Factors , Demonstrate statistical equivalence or better INPUTS OUTPUTS PROCESS Quality Process Characteristics Factors Table A: Table B: Describe the Measures of Change Process/Product Noise Table C: Performance Factors Range of Application of the Change Change Control for Materials MTL008986 Rev 3.1 (CSMO) May 2002 16 White Paper – Notes Purposes of Tables A, B, C and D in the White Paper 1. Use as a tool to develop the qualification plan. 2. Present all variables in a standardized format that can be clearly understood and discussed. 3. Identify all the applicable variables: , Process Factors (Table A) , Quality Characteristics (Table B) , Noise Factors (Table C) 4. Identify which variables will be evaluated. 5. Document how the variables are measured. 6. Identify potential interactions between variables. 7. Describe the experimental design (Table D). 8. Define the sampling plan and sample sizes. 9. Define the acceptance criteria. Change Control for Materials MTL008986 Rev 3.1 (CSMO) May 2002 17 White Paper – Notes Table A – Process Factors Example: Process Factor Present Value Proposed Value Cure Time 60 minutes 15 minutes Acid Concentration 10% 20% Substrate Length 5.0 inches 4.8 inches Alignment Mark Design Cross Circular Mold 4 cavity 8 cavity Resin Supplier AAA Resin Company Top Resin Company Manufacturing Location Columbia, Missouri San Diego, California Change Control for Materials MTL008986 Rev 3.1 (CSMO) May 2002 18 White Paper – Notes Table B – Quality Characteristics Example: Quality Measurement Ideal Present Present Expected Characteristic Method Target Mean Std Dev Change Electrical Shorts HP Tester 0% 3.5% NA Improve Solder Thickness Cross Section 1 micron 10.0 2.0 Decrease microns microns mean Outgoing mechanical Visual Inspection 0 dpm 50 dpm NA None defects Viscosity Viscometer 10000 kcps 10500 800 kcps None kcps Heat Slug Fillet Voids X-Ray 0/part 5/part NA Improve EXERCISE 4 Please refer to Exercise 4 in the Student Handout. Change Control for Materials MTL008986 Rev 3.1 (CSMO) May 2002 19 White Paper – Notes Table C – Noise Factors Example: Noise Factor Range of Values/Scope Range Evaluated in Experimental Strategy Solder Rosin Lots 3000 units per lot 8 lots Production lines at ABC 3 lines 2 lines Heat slug lots 15,000 pcs/lot 2,000 pcs/lot Shifts at ABC assembly 2 shifts 2 shifts AAA Solder Company Solder supplier BBB Solder Inc. BBB Solder Inc. Operators 6 per shift 2 per shift Manufacturing line 10 lines #1, #10 Change Control for Materials MTL008986 Rev 3.1 (CSMO) May 2002 20 COMPLETE TABLES A THROUGH D FOR PWP DO NOT CHANGE FOR FWP TABLE D EXPERIMENTAL STRATEGY Exp # Exp Material Quantity Process Factors Noise Factors Quality Accept Criteria (Table A) (Table C) Characteristics (Table B) Summarize experimental plan by listing out each experiment conducted. There may be experiments conducted at the supplier site and/or the Intel site. Details of each experiment will be documented on the following pages. The “Accept Criteria” column is used to indicate if the quality characteristics have to remain the same as current values, or if the point of the proposed change is to shift those values. For example, if the proposed change is to improve yield, then the accept criteria will be a statistically or technically significant increase in yield. Accept Criteria: ， SE = Statistically Equivalent, based on statistical test of data ， SB = Statistically Better, based on statistical test of data ， SEB = Statistically Equivalent or Better ， Capable = Cpk > 1.33 ， Stable = Process is stable ， TE = Technically equivalent ， GRC = Gross Reality Check. Should only be used when sample sizes are extremely large, or there is a very low risk. Change Control for Materials MTL008986 Rev 2.1.2 (CSMO) May 2002 21 COMPLETE TABLES A THROUGH D FOR PWP DO NOT CHANGE FOR FWP TABLE D (continued) Detailed Description Experiment Purpose: To document the details of each experiment. To provide sufficient details to conduct the experiment. Provide details on: 1. Objective of the experiment: Document what decisions this data will be used to support – Capability, stability, etc. 2. Choice of experimental material and justification: Indicate test vehicle, live product, part number if change applies to many parts 3. Experiment design: paired sample retest two sample serial paired sample split lot one sample two sample cross-over zero defect sample two sample concurrent other __________________________ 4. Sample size justification for each quality characteristic: Use this table to document the minimum sample size statistically required to detect the stated technically significant delta. If more or less samples will be used, indicate this in the “n actual” column and briefly explain why. Quality Characteristic D n n , , , Req’d Actual 5. Sample selection Explain sampling plan to satisfy n required 6. Experiment/material flow Explain flow for how and where material will be built and tested, insert a flow diagram on next page if required. Change Control for Materials MTL008986 Rev 2.1.2 (CSMO) May 2002 22 7. Statistical analysis methods ， N/A if GRC is used. ， One sample t-test/binomial (for one sample) ， Two sample t-test/binomial (for two sample cross-over, two sample concurrent or two sample serial) ， Paired t-test (for paired sample retest) ， ANOVA t-test or k-sample binomial (for more than two groups) ， Capability analysis if Cpk is the accept criteria ， Control charts if stability is the accept criteria Insert a flow diagram for experimental flow here, if required. Delete this page if not needed. Use this page to show how material will be built, where samples are pulled, if lots are split, where samples are shipped. This information will also be useful to those responsible for executing the experimental plan. Change Control for Materials MTL008986 Rev 2.1.2 (CSMO) May 2002 23 White Paper – Notes EXAMPLE: TABLE D Exp Exp Material Quantity Process Noise Quality Characteristics Accept Criteria # Factors Factors (Table B) (Table A) (Table C) 1 3 Jones solder 2.25g/lot All # Solder Ionic Content of Solder Paste For paste lots, 2 Smith Rosin Information solder paste lots Lots Only 2 3 rosin lots Jones ~6000 All All Solder Flux Residue Defect Level, SEB solder paste, piece Heat Slug Delamination, ABC 4331 PPGA parts Chip Capacitor Solder Fillet Height, Solder Fillet Defects, Heat Slug Fillet Voids, Flux Residue Signature Prior to Bake, Flux Residue Signature Post Bake Change Control for Materials MTL008986 Rev 2.1.2 (CSMO) May 2002 24 White Paper – Notes Experimental Strategy Guidelines – Design Efficiency There are many considerations in selecting an experimental design. A balance must be achieved between statistical efficiency and practical logistics. Experimental designs are listed below in order of statistical efficiency. Start with the first and move down until a practical experimental design is identified. Paired Sample Retest: all units receive both treatments Paired Sample Retest - Experimental Flow, Works well for metrologies or test condition changes. Time, Raw data is required from both processes to perform the analysis. UnitsRandomize order ofAdvantages application, Removes any unit-to-unit variation from the comparison. Process 1, Lower sample sizes required for same (POR or New)detectable delta vs. two sample comparisons. , No assumptions about current baseline Process 2needed. (POR or New) Disadvantages Compare , Can only be used on process that don’t permanently alter the experimental units. Change Control for Materials MTL008986 Rev 2.1.2 (CSMO) May 2002 25 Paired Sample Split Lot: each lot is split at the proposed process factor change such that Paired Sample Split Lot - Experimental Flow? the lot receives the current treatment and the other ? receives the proposed treatment , Example: each lot is split between two Timeplating baths. Lot, Raw data is required from both processes to perform the analysis. Advantages , Removes any lot-to-lot variation from the analysis with the assumption that the units within the lot are essentially identical as far as responding to the applied treatment Process 1Process 2, Lower sample sizes required for same POR(New)Comparedetectable delta vs. two sample comparisons. , No assumptions about current baseline needed. LotDisadvantages , Difficult to administer , Non-standard process flow. Change Control for Materials MTL008986 Rev 2.1.2 (CSMO) May 2002 26 Two Sample Cross-Over: both treatments are run concurrently then machines are switched so that all machines eventually run all Two Sample Cross-Over - Experimental Flowtreatments. , Example: old and new cure profile run on each of two ovens. LotLotTime, Raw data is required from both processes to perform the analysis. Machine #1Machine #2Advantages , Easier to administer in the factory Process 2Process 1, Can be run as part of the standard process (New)(POR)flow. , By crossing over, removes any effect from Comparethe machines from the analysis. , By running concurrently, confounding time Machine #1Machine #2effects are minimized. No assumptions about current baseline needed. Process 1Process 2 (POR)(New)Disadvantages , Requires somewhat larger sample size than the paired tests listed above Compare, Requires careful sample selection to ensure that results aren’t biased by differences in experimental material. Change Control for Materials MTL008986 Rev 2.1.2 (CSMO) May 2002 27 Two Sample Concurrent: both treatments are run concurrently; no swapping. , Example: qualifying a second line. Two Sample Concurrent - Experimental Flow, Raw data is required from both processes to perform the analysis. TimeAdvantages , Easier to administer in the factory , Can be run as part of the standard process flow. LotLot, By running concurrently, confounding time effects are minimized. , No assumptions about current baseline Machine #1Machine #2needed. Process 2 Process 1 (New)(POR)Disadvantages , Requires somewhat larger sample size than the paired tests listed above , Requires careful sample selection to Compareensure that results aren’t biased by differences in experimental material , Allows for machine-to-machine difference to possibly confound results by not crossing over. Change Control for Materials MTL008986 Rev 2.1.2 (CSMO) May 2002 28 Two Sample Serial Pilot: one treatment is run followed by the second treatment. Two Sample Serial - Experimental Flow , Example: convert machine to a new recipe or process. , Raw data is required from both processes Time Lot to perform the analysis. Advantages , Easier to administer in the factory Machine , Can be run as part of the standard process #1 flow. Process , No assumptions about current baseline 1 (POR) needed. Compare Disadvantages , Requires somewhat larger sample size Machine than the paired tests listed above #1 , Requires careful sample selection to Process ensure that results aren’t biased by 2 (New) differences in experimental material , Allows for confounding time effects due to not running concurrently. Factory excursions can bias or invalidate results unbeknownst to the experimenter Change Control for Materials MTL008986 Rev 2.1.2 (CSMO) May 2002 29 One Sample: the proposed treatment condition is run and the results compared to a fixed baseline value like the design target or the current process mean. , Example: new process is compared to existing CofC data. One Sample - Experimental Flow, Raw data is only required from the new process to perform the analysis. Advantages , Easy to administer in the factory Time , Can be run as part of the standard process flow , Requires smaller sample sizes than two sample comparisons. Disadvantages Lot, Requires careful sample selection to Compare toensure that results aren’t biased by differences in experimental material that baseline receive the treatment vs. the baseline data Machine #1, Allows for machine-to-machine difference to possibly confound results by not Process crossing over. (New), Allows for confounding time effects due to not running concurrently. , Factory excursions can bias or invalidate results unbeknownst to the experimenter. , Baseline data must be substantial, stable, and well documented. , Risky approach if the historical baseline has significant variation or is unstable. Including relevant sources of noise is critical. Change Control for Materials MTL008986 Rev 2.1.2 (CSMO) May 2002 30 Zero Defect Sample: proposed treatment condition is run and no failures are allowed , A plan which typically requires that the actual yield loss or failure rate be much less than the level we want to demonstrate , If you expect failures, then do not use zero defects sampling Advantages , Requires much lower sample size than one or two sample binomial comparisons , Simple analysis, pass on zero defects, fail on more than zero. Disadvantages , Requires that the actual defect level be substantially lower than the upper confidence limit. Experiments should run a control leg (i.e., two sample comparisons) whenever possible. Only use a one sample test in the following situations: , It can be shown that the target value from historical data is stable. , You are matching the first machine/process/material of its kind to a target value. The second instance of this change should be matched to the current values of the first one, using a two sample comparison. , Zero defect sampling. Zero defect sampling is not suitable for comparing 2 processes. Statistics training classes contain documentation of Intel’s best known methods for designing and analyzing data. These BKM’s should be followed whenever possible, and justification provided when a deviation occurs. Change Control for Materials MTL008986 Rev 2.1.2 (CSMO) May 2002 31 White Paper – Notes Experimental Strategy Flow Experimental Strategy Flow Table to Complete List Process Factors to be Table A changed. , List Quality Characteristics Table B affected by Process Factors. , Add Noise Factors, which may Table C affect the Quality Characteristics. , List the sequence of Table D experiments that define the Experimental Strategy. , Attach a complete description of Table D each experiment. Change Control for Materials MTL008986 Rev 2.1.2 (CSMO) May 2002 32 STOP AT THIS POINT WHEN COMPLETING YOUR PRELIMINARY WHITE PAPER Quality Characteristics Summary Quality Characteristic Ideal POR POR New New , Target Mean Std Dev Mean Std Dev Result (TSD) or Spec The first 5 columns in this table will come from information previously documented in the PWP. Add it here and include the new data from the experiments. Results can be indicated as: ， SE = Statistically Equivalent, based on statistical test of data ， SB = Statistically Better, based on statistical test of data ， SEB = Statistically Equivalent or Better ， TE = Technically equivalent ， Pass or Fail = Passes or fails Gross Reality Check (GRC) criteria. Pass or fail are also used when accept criteria is capability and/or stability ， NE = Not Equivalent, data is not SE or TE – provide justification for implementation of change if still planned. Change Control for Materials MTL008986 Rev 2.1.2 (CSMO) May 2002 33 Conclusions: Summarize key results and implications of the results Recommendations: Document change team’s recommendations for implementation, rejection, additional data collection, needed follow-up, etc. Include issues/implications, team’s recommendation. Other Implementation Items / Details/SLI’s: Document follow-up plans for implementation, what SLI’s or Lot numbers this change will be implemented on. Results Details: (Attach supporting analyses and graphs here) Change Control for Materials MTL008986 Rev 2.1.2 (CSMO) May 2002 34 White Paper – Notes Final White Paper Purpose: 1. Document the results of the experiments. 2. Provide a conclusion as to whether the change is equivalent, better or worse than the current process. 3. Provide a recommendation: , Implement , Do not implement , Collect more data 4. Provide historical documentation for future reference. 5. Use as a tool to "sell" customer on needed or proposed changes. Change Control for Materials MTL008986 Rev 2.1.2 (CSMO) May 2002 35 White Paper – Notes Final White Paper – Statistical Analysis Guidelines All experimental data should be plotted and statistically analyzed using the methods documented in the Intel statistical curriculum. A complete statistical analysis should include: , Key structure plots, which show the full structure of the raw data. , Noise factors such as date and machines should be represented in the plots. , Key plots and tables of summarized data. , Estimates of differences. , 95% confidence intervals for differences and effects preferable to t values and significance levels. Change Control for Materials MTL008986 Rev 2.1.2 (CSMO) May 2002 36 Exercise 5 Please refer to Exercise 5 in the Student Handout. Change Control for Materials MTL008986 Rev 2.1.2 (CSMO) May 2002 37