Fluent Essays

Module 4 AssignmentIn this assignment, you will apply what youve learned in this module about the designs of experiments to a sample data set and scenario.Assignment InstructionsConsider the following: The ABC Company wants to optimize the response (click rate) to their online ads. After a brainstorming session, thirteen factors were identified as potentially having an effect on the response (click) rate. The table below lists the 13 factors and the two levels that should be considered. This data is available in the DOE Assignment JMP file attached below. Identified Response FactorsTeaser OfferTelephoneNumberGraphicFont SizeAdvertisingChanelMessageTypeHeadlineLayoutProductselectionGift OfferProduc InfoColorSchemaDiscountNumber ofClicksLevel 1YesYesYesLargeYesAHeadline 1StandardFeature AYesVersion AAYesLevel 2NoNoNoSmallNoBHeadline 2CreativeFeature BNoVersion BBNo1YesYesYesLargeYesAHeadline 1StandardFeature AYesVersion AAYes522NoYesNoLargeYesBHeadline2CreativeFeature AYesVersion BANo383YesNoNoSmallYesBHeadline 1StandardFeature BYesVersion BBNo424YesYesYesSmallNoBHeadline 1CreativeFeature ANoVersion BBYes1345NoYesYesLargeNoBHeadline 1CreativeFeature BYesVersion BBYes1046NoNoNoLargeYesAHeadline 1CreativeFeature ANoVersion ABYes607NoNoYesSmallYesAHeadline 2CreativeFeature BYesVersion BAYes618NoNoYesLargeNoBHeadline 2StandardFeature BNoVersion ABNo689YesNoNoLargeYesBHeadline 1StandardFeature BNoVersion BAYes5710NoYesNoSmallYesAHeadline 1CreativeFeature BNoVersion ABNo3011YesNoNoSmallNoBHeadline 2CreativeFeature ANoVersion AAYes10812NoYesNoSmallNoBHeadline 1StandardFeature AYesVersion AANo3913YesNoYesSmallNoAHeadline 1CreativeFeature BYesVersion AANo4014YesYesNoLargeNoAHeadline 2CreativeFeature BNoVersion BANo4915YesYesYesSmallYesAHeadline 2StandardFeature ANoVersion BBNo3716YesYesNoLargeNoAHeadline 2StandardFeature BYesVersion ABYes9917NoYesYesSmallYesBHeadline 2StandardFeature BNoVersion AAYes8618NoNoYesLargeNoAHeadline 1StandardFeature ANoVersion BANo4319YesNoYesLargeYesBHeadline 2CreativeFeature AYesVersion ABNo4720NoNoNoSmallNoAHeadline 2StandardFeature AYesVersion BBYes104Discuss how you approach the problems and answer the questions along the way.1. How many treatments do you need at a minimum to estimate 13 main effects and the overall mean? Find a design using JMP DOE>Classical Designs>Screening Designs add 13 factors and find a design.What is the minimum number of treatments (runs)?What is the fractional factorial design with the smallest number of runs you can use for our problem?Which other design could you choose?2. Once you have a design matrix you would carry out the treatments and collect the response for each treatment (run). To exercise the analysis of a design I provided you with a 20 run design (which is the Packett-Burman design shown ion your list) which has responses provided in the JMP file. Use the design matrix provided in the DOE_Assignment_5_Click(2).jmp file with the number of clicks as the response variable. Use DOE>Classical>Two Level Screening> Fit Two Level Screening . Which factors are statistically significant at p<=0.05? Highlight the statistically significant factor rows (use individual p) and click run model. Interpret the output. 3. To evaluate the current design matrix with just the few main factors you determined to be significant, use DOE>Design Diagnostics>Evaluate Design. Select only the statistically significant factors for the evaluation. Look at the Alias Matrix to see what the problems are with using the same 20 runs to estimate the interaction effects (confounding of main effects and interaction effects). Specifically, we are interested in the 2-factor interaction between the biggest effects. What main factors are confounded with this interaction and what is the magnitude? Interpret the finding. (Note: find the column of the 2-factor interaction which had the largest effect. Then see what row has a number different from zero and what the main effect in that row is. The larger the absolute value the larger is the confounding. )4. Now we want to evaluate the data based on our discovery that the 2-factor interaction is confounded with another important factor. Go back to the open window you had before (DOE>Classical>Two Level Screening> Fit Two Level Screening ) and select the statistically significant factors plus the 2-factor interaction of the two biggest effects. Click Run Model again. Interpret the output. What is likely happening?5. Make a final selection on the window (DOE>Classical>Two Level Screening> Fit Two Level Screening) of what you think is the true likely factors and or interactions. Then click Run Model again. Interpret final model. Prepare the report using the following formatting guidelines:1 page, single-spaced report using 0.5 margins and two-column format1 page for appendixInclude title of report, then FirstName, LastName, ISDS course #, Assignment #, date (00/00/00)10 pt Font Calibri or Times New RomanJustified as sample reportCreate headings for each sectionList any references used (e.g. Module 1 Resources)Include a title for your report e.g. Text Analysis of Workers Compensation Claims and create headings for each sectionInclude supporting relevant figures from the analysis in your AppendixSubmit as pdf with filename first name initial last name and assignment number (for instance HSchneider#1)Be sure to review the Assignment Rubric and Assignment Example attached below. If you have any questions, please post in the Module Questions Forum. assignment__4_example.pdf assignment__4_guide.pdf assignment__4_rubric.pdf isds_7303_module_3_7_assignment_rubric.pdf Unformatted Attachment Preview Name Assignment 4 Example Introduction and Business Understanding A credit card company sent offers to potential customers as a way to increase revenue with a goal of optimizing the customer response rate. Although they identified many variables, one challenge was understanding which were the most important to optimizing the response rate. To support this optimization, Design of Experiment (DOE) was used to determine which factors were important. Data Understanding: Nineteen two-level factors were identified by the credit card company as possibly having an impact on customer Response Rate, which was the continuous response variable. The full list is in Appendix A. All two-level factors in the file were treated as nominal predictor variables. Experimental Designs: Several DOE designs were considered using all 19 factors with the Screening Designs platform under Classical Designs option. A full factorial design for the 19 factors, would require 219 runs. This was not practical for an experiment and therefore not evaluated. JMP provided multiple options to reduce the design to a fractional factorial, including minimum runs, default runs, and user selection. The minimum represented the number of factors plus one, which was 20. The default provided a balanced design with more than the minimum, but less than the full factorial. The default design was 24 runs. Both the 20 and 24 run options were evaluated in the Distribution platform and it was determined that a 20 runs would be less costly and also provided a balanced design. Identifying Significant Factors: After selecting a design with 20 runs, the data file was used to identify the statistically significant factors at a significance level of p= 0.05. Fit Two Level Screening under the DOE platform was used to analyze the design. Three factors (interest rate, sticker and SECOND_BUCKSLIP) were statistically significant at 0.05, but COPY_MESSAGE was borderline statistically significant (p=0.0512). The remaining 15 factors had p-values greater than 0.16. Next the four biggest effects were selected and the model run (see Appendix C). All four factors are statistically significant at P<0.0075. Confounding: Once the statistically significant variables were identified, their two factor interactions were evaluated using the Alias Matrix under Evaluate Design BADM 7020 platform. This matrix showed the degree of bias imparted to model parameters1. The largest entries in the alias matrix were two-factor interactions between INTEREST_RATE, STICKER, and SECOND_BUCKSLIP at a value of -0.6, indicating these three factors are confounded and that the magnitude of potential bias is -0.6 of the interaction effect. The remaining 3 factors all had two-way interactions with each other at a bias of -0.2. See Appendix D for the Alias Matrix table. Reduced Design: Based on the findings from the Alias Matrix a new model was created using Make Model in JMP. This model included the four factors identified plus the interaction between Sticker and interest rate. The result (Appendix E) shows that the SECOND_BUCKSLIP and interaction interest rate and sticker are not simultaneously statistically significant at p<=0.05. It is likely that the interaction between sticker and interest rate is truly different from zero rather than the SECOND_BUCKSLIP. Therefore, the SECOND_BUCKSLIP was removed and the model rerun. The result shows that the three factors INTEREST_RATE, STICKER, COPY_MESSAGE and the interaction between STICKER and INTEREST_RATE are all statistically significant at p<0.027. The Interaction Plot (Appendix G) shows that the effect of interest rate is larger for when there is a sticker than no sticker. Conclusion: When attempting to determine important factors for customer response rate to credit card offers, Design of Experiment was employed. Fractional factorial design was used to limit the number of runs which would lower cost while still providing enough runs to estimate. The first screening design was able to identify 4 of the 19 factors to be statistically significant. However, the Alias Matrix showed that there was confounding between the 2factor interaction of the two largest factors and SECOND_BUCKSLIP. An analysis of the design with the four factors plus the interaction showed that most likely model included the three factors and the two factor interaction of the largest effects. The highest response rate was achieved for low interest rate, a sticker and a generic copy message. The strength of the design was that with just 20 runs 19 factors could be evaluated. However, a second experiment with just 8 runs including the four factors might be necessary to confirm that the 2-fcator interaction rather than the SECOND_BUCKSLIP is the important factor, because of the confounding. Appendix C: Parameter Estimates for 4 Biggest Effects Appendix D – Alias Matrix for 4 Statistically Significant Factors and 2-way Interactions Appendix E: Result of Reduced Design Appendix B: Design Analysis Result Appendix F: Final Model Appendix G: Interaction Plot Step 1: Use DOE> Classical > Screening Designs Replace the 1 with 13 and continue. Add 13 categorical factors needed for the design. If there are any other factors already there then delete them. After continue you will get a list of designs. (Note, your list will be wore extensive.) Step 2: Use DOE>Classical>Two Level Screening> Fit Two Level Screening Select all factors as X variable and Number of Clicks as Y variable. (Note: you do not include Treatment, which is not a factor!) Click OK. Select (highlight) all the statistically significant factors and click Run Model. Interpret the output. Step 3: Use DOE>Design Diagnostics>Evaluate Design to evaluate the design. The rule of thumb in experimental designs is that if the main effect is not large then the interactions are not large either. Therefore select only statistically significant Factors into X and Number of Clicks into Y. You get a lengthy output but we are only interested in the Alias Matrix. The Alias Matrix you will see has your main effects in the first column and the interactions in the first row. The matrix entries tell you about the magnitude of confounding. The large the absolute value the larger the confounding between the main effect and the interaction effect. Ignore the intercept. For instance if your first main effect listed is advertising Chanel and the first interaction between Advertising Chanel and Message Type is zero as shown in the figure above then there is no confounding between these two. If it were greater than zero or smaller than zero than there is confounding. Following the rule of thumb again we are likely to have an interaction between two factors important when their main effects are large. Therefore look specifically at the interaction of the two largest main effects and which main factor this interaction is confounded with. This is where your main confounding is. Step 4: Run DOE>Classical>Two Level Screening> Fit Two Level Screening) again with adding the interaction you identified in Step 3 You have to highlight this interaction in addition to your statistically significant main effects. Interpret the output. What could likely be happening? Step 5: Make a final selection of main effects you think are truly different from zero with the appropriate interaction and run DOE>Classical>Two Level Screening> Fit Two Level Screening again. The interaction plots can be obtain the Factor Profiling> Interaction Plot. They look a little different but provide the same information as the one shown in the sample assignment. Interpret the output. Write your report using the sample report as guide. Introduction & Business Understanding Data Understanding Analysis Design Analysis Factors Confounding Design Follow-Up Conclusion Format and Overall Writing. Provides background on the business problem. Explains clearly the issue of issues of DOEs. 10points Does demonstrate some understanding of the business issues. But explanation of purpose of DOE vague or unclear. 7points Describes the general business issue at hand. Lacking specificity or clarity about DOE. 5points No business Understanding. 0points Demonstrates superior data understanding as it relates to the specific model. 10points Demonstrates understanding of data issues but misses some important details. 7points Data understanding not sufficiently described. Major details are missing. 5points No data understanding not sufficient. 0points Selects correct design. 10points Selects adequate design. But design not optimal. 7points Design not adequate. 5points No design. 0points Identifies significant factors. 10points Identifies significant factors, but list is missing a variable. 7points Analysis not complete. 5points Very poor analysis. 0points Identifies confounding and explains it correctly. 10points Identifies confounding but explanation is incomplete or incorrect. 7points Confounding not explained adequately. 10points 5points No explanation of confounding. 0points Creates correct followup fractional factorial design with minimum runs necessary. 10points Follow up design is not optimal. 7points Follow up design is insufficient. 5points Follow up design missing. 0points Conclusion clearly describes the findings and identifies the strength and weakness of the designs used. 20points Conclusion describes findings, but does not include a clear description of strength and weakness of designs. 17points Conclusion is too vague and general. 15points Conclusion missing. 0points Follows format. Text has superior layout with headings. Writing is style specific and detailed and professional. Follows format, and average writing style with some minor issues. 7points Does not follow required format or writing style is not adequate for a professional report. 5points Not formatted writing style unprofessional. 0points 10points Appendix Appendix has meaningful charts, numbered and titles. Has all charts required for understanding. 10points Has appendix but some tables/graphs are missing or labeled wrong. 7points Many graphs and titles are missing. 5points No ISDS 7303 Module 3-7 Assignment Rubric The following rubric will be used to grade the assignments for Modules 3-7. Introduction & Business Understanding Description demonstrates understanding of business issue at hand. Explains specific issues to be addressed in clear and specific language. 10 points Describes the general business issue at hand. Lacking specificity or clarity 5 points Does not demonstrate understanding of the business issues. Explanation is vague or unclear. 0 points Data Understanding & Description demonstrates Preparation superior data preparation and understanding. Utilizes appropriate recoding based on relationship between inputs & outputs, addresses missing values using advanced methods and makes transformations as necessary. 20 points Demonstrates understanding of data issues. Prepared data sufficiently. 10 points Data understanding not sufficient. Data preparation not sufficient. 0 points Analysis Does not demonstrate understanding of the business issues. Explanation is vague or unclear. 40 points Provides adequate preparation of data with some analysis and coding. 20 points No adequate solution of the problem. Data preparation not sufficient. 0 points Conclusion Conclusion recaps issues and addresses implementation with specific insights. Provides specific recommendations derived from the model. 20 points Conclusion does provide a recap of the issues and addresses implementation. 10 points Conclusion is too broad. Does not address specific issues of the case. 0 points Format Follows report format Follows format mostly with exactly. Text has superior some minor issues layout with headings. 5 points Appendix has meaningful charts, numbered and titles. References graphs and tables in text and clearly explains them to support findings.10 points Does not follow required format. 0 points ... Purchase answer to see full attachment