| Course | GB730 Applied Business Research Design and Methodology |
|---|---|
| Unit | Unit 9 |
| Paper type | Data analysis plan |
| Length | About 1,160 words, 7 pages double-spaced |
| Format | APA 7 student paper |
| School | Purdue University Global |
| Program | MBA |
| Updated | October 2026 |
Free sample paper for GB730 Unit 9
Decisions Fixed Before the Data: A Unit 9 Data Analysis Plan for Hotel Rate Overrides
[Student Name]
Purdue University Global
GB730: Applied Business Research Design and Methodology
Unit 9 Data Analysis Plan
[Instructor Name]
[Date]
The hotel company and the study are composite and illustrative, and no data have been collected; the methods sources cited are real.
Why the Plan Comes First
An analysis chosen after the data are seen can be shaped, knowingly or not, by what the data show. A researcher who tries several models and reports the one with the clearest result has, in effect, tested many hypotheses and reported one. Writing the plan now, and filing it with the dissertation chair before extraction, commits the study to a set of choices that can be checked later. Any departure from the plan will be reported as a departure, with its reason.
Preparing the Data
Four rules are set now. First, duplicates: a hotel-night appearing twice in the property management extract, which happens when a night audit is rerun, will be resolved by keeping the later record. Second, missing competitor rates, which the Unit 7 memo flagged on some holiday nights, will be handled by multiple imputation using the hotel, the date and the hotel's own rate, with twenty imputed datasets; a sensitivity analysis will repeat the main models on complete cases only. Third, hotel-nights when a hotel was closed or had more than half its rooms out of order will be excluded, since occupancy on those nights does not reflect pricing. Fourth, the first three months after each hotel's system went live in 2022 fall outside the study period already, so no settling-in period needs to be removed.
Research Question 1: Overrides and Outcomes
The predictor. Override status will be coded in three categories rather than two: no override, an override above the recommended rate and an override below it. Treating upward and downward overrides as one category would hide the most important pattern the study might find, since a manager who raises rates before a sold-out event and a manager who cuts rates on a slow Tuesday are doing opposite things.
Occupancy. Occupancy will be modeled with a linear mixed model, with random intercepts for hotel and for manager, since each hotel has its own baseline and each manager may have their own style. The adjustment variables named in Unit 7 enter as fixed effects: bookings on hand at fourteen days, the occupancy the system itself forecast two weeks out, competitors' median rate, the event flag, group rooms blocked, day of week and month. The coefficient of interest is the difference in occupancy, in percentage points, between nights with each type of override and nights without one, reported with a 95 percent confidence interval.
Average daily rate. Average daily rate will be modeled the same way, using its logarithm so that coefficients can be read as approximate percentage differences. Rates differ widely between a select-service hotel in a small city and a full-service hotel downtown, and a percentage reads more meaningfully across that range than a dollar figure.
Checking the models. Residuals will be inspected for patterns over time, since nights close together in the calendar are likely to resemble one another. If they show serial dependence, standard errors will be clustered by hotel and month as a sensitivity check. No model will be chosen on the basis of which produces significant results.
How Much Unmeasured Knowledge Would It Take?
The Unit 1 post identified the study's central weakness: managers override when they know things the records do not show, and that knowledge affects outcomes directly. Adjustment cannot remove it. The plan therefore includes an E-value for each main association, following VanderWeele and Ding (2017). The E-value states how strongly an unmeasured factor would have to be associated with both overriding and the outcome to fully explain away the observed association. A small E-value means a modest unmeasured factor could account for the finding; a large one means it would take a strong one. This does not solve the selection problem, but it states its size in a way readers can judge.
The E-value does not tell the reader whether managers know something the data miss; it tells them how much they would need to know to explain the result away.
Selecting Interviewees From the Results
The connecting step described in Unit 4 depends on characterizing each manager's overrides. After the main models are run, each manager's own association between overriding and outcomes will be estimated by adding random slopes for override status by manager. Managers will then be sorted into three groups: those whose overrides associate with higher rates without lower occupancy, those whose overrides associate with lower occupancy, and those whose overrides show little association either way. About five managers will be invited from each group, through the information systems director, as the Unit 5 plan describes.
Research Question 2: Managers' Accounts
For the interviews the plan adopts reflexive thematic analysis (Braun & Clarke, 2006). The candidate will read every transcript twice before coding, code line by line, gather related codes into provisional themes, check each provisional theme against the whole interview set, settle each theme's name and scope, and only then write the findings. Braun and Clarke (2019) stress that themes are built by the researcher rather than found waiting in the data, which is why the candidate's position must be documented alongside them. The reflexive approach suits a study in which the candidate's own position is part of the analysis rather than something to be eliminated, and the reflexive journal from Unit 8 will be read alongside the coding. Codes will be generated inductively from the transcripts, not from a list drawn up in advance, so that managers' own categories of reasoning can emerge.
Research Question 3: Bringing the Strands Together
The strands will be joined in a joint display, which places quantitative and qualitative results side by side so that their relationship can be seen (Guetterman et al., 2015). The rows will be the three groups of managers. The columns will give each group's quantitative pattern, the themes that appeared most often in that group's interviews, and a judgment of whether the accounts confirm, expand on or contradict the pattern. A group whose overrides associate with higher rates might, for example, describe acting on information about local events the system does not receive, which would expand on the quantitative result. A group whose overrides associate with lower occupancy might describe the same reasoning, which would contradict it and raise a different question.
What the Plan Will Report Regardless of Results
The study will report every model named here, including those with null results, and the sensitivity analyses alongside the main ones. If the main associations are near zero, the study will say so, and the interviews will be used to explore why managers believe overrides help when the records show little difference.
References
Braun, V., & Clarke, V. (2006). Using thematic analysis in psychology. Qualitative Research in Psychology, 3(2), 77-101. https://doi.org/10.1191/1478088706qp063oa
Braun, V., & Clarke, V. (2019). Reflecting on reflexive thematic analysis. Qualitative Research in Sport, Exercise and Health, 11(4), 589-597. https://doi.org/10.1080/2159676X.2019.1628806
Guetterman, T. C., Fetters, M. D., & Creswell, J. W. (2015). Integrating quantitative and qualitative results in health science mixed methods research through joint displays. Annals of Family Medicine, 13(6), 554-561. https://doi.org/10.1370/afm.1865
VanderWeele, T. J., & Ding, P. (2017). Sensitivity analysis in observational research: Introducing the E-value. Annals of Internal Medicine, 167(4), 268-274. https://doi.org/10.7326/M16-2607
The Purdue Global GB730 Unit 9 instructions
Unit 9 in GB730 wants the analysis for every research question settled before any data are collected. Explain why the plan is fixed in advance. Set out data preparation rules for duplicates, missing values and exclusions. For each quantitative question, name the model, the outcome, the predictor and its coding, the adjustment variables and how results will be reported, with intervals. Describe how you will check model assumptions and what you will do if they fail. Address the main threat to your inference with a stated sensitivity analysis. In a sequential design, explain how results will select the qualitative participants. Name and describe the qualitative approach and its phases. Describe how the strands will be integrated. Close with what you will report whatever the results. Eight pages organized by research question is typical.
How this GB730 Unit 9 example is built
The plan opens by explaining why choosing models after seeing data amounts to testing many hypotheses. Four preparation rules follow, including twenty imputations for missing competitor rates with a complete-case check. Question one gets a three-category predictor, mixed models for occupancy and log rate, the adjustment set from Unit 7 and a residual check for serial dependence. A section on unmeasured knowledge introduces VanderWeele and Ding's E-value, with the highlighted sentence explaining what it does and does not say. Random slopes by manager sort managers into three groups for interviews. The qualitative section follows Braun and Clarke's reflexive approach in plain steps, and the joint display section imagines both confirming and contradicting rows. The plan commits to reporting null results and every sensitivity model.
GB730 Unit 9 rubric: what earns full points
The GB730 analysis plan is graded on whether another researcher could run it. Instructors give heavy weight to complete model specifications: outcome, predictor coding, adjustment variables, random effects and reporting format. Preparation rules for duplicates, missing data and exclusions are often a separate rubric row. A named sensitivity analysis for the study's main threat earns credit because it shows the student knows where the inference is weakest. For sequential designs, the rule that turns results into interview selection must be stated in advance. The qualitative approach should be named and its steps described, not just cited. An integration method, such as a joint display, completes the mixed design. Committing to report null findings shows research integrity and is noticed in this course.
Common GB730 Unit 9 mistakes, and how to avoid them
Many GB730 analysis plans name a test, such as regression, without saying which variables enter or how the predictor is coded. Missing data are often left for later, which means the decision will be made after seeing results. Students frequently treat a categorical predictor as two groups when the direction of the effect matters. The main threat to inference is described in the design chapter and then forgotten in the analysis. Qualitative plans cite Braun and Clarke without describing what will actually be done with transcripts. Integration is sometimes reduced to discussing both strands together. Specify every model fully, write preparation rules, code the predictor to match the question, size the main threat, describe the qualitative steps and plan the joint display. Send us your questions and data fields for a plan written to them.
Get a GB730 Unit 9 example written to your instructions
Send research questions, instruments and design, plus the Unit 9 analysis plan requirements and the rubric they sit under. In 24-48h a custom plan comes back mapping each question to a single analysis justified by the data's structure, with preparation rules and the integration step set out. First orders are never invoiced. The paper above is an original model document written by our desk, not a submitted student paper and not an official Purdue University Global document.
More GB730 papers and MBA samples
- GB730 Unit 1: Discussion Board Post
- GB730 Unit 2: Problem and Purpose Memo
- GB730 Unit 3: Question Alignment Exercise
- GB730 Unit 4: Design Justification
- GB730 Unit 5: Site Access Plan
- GB730 Unit 6: Seminar Reflection
- GB730 Unit 7: Instrument Development Memo
- GB730 Unit 8: Qualitative Rigor Plan
- GB730 Unit 10: Methodology Chapter Draft
- GB720 Unit 9: Integrative Framework Paper
- GB710 Unit 9: Research Problem Statement
- GB701 Unit 9: Assumption Diagnostics Memo
- GB700 Unit 9: Problem Statement Draft
GB730 Unit 9 questions, answered
How detailed should a GB730 analysis plan be?
Detailed enough that another researcher could run the analysis without asking the author a single question. Name the procedure, the software, the variables in each role, the handling of missing data and the decision rules. For qualitative analysis, name the approach and describe its phases. Vague plans are commonly returned because nothing in them can be checked later.
Does the plan need a power analysis?
Where sample size is a real constraint, yes. With large administrative datasets power is often ample, and the plan should say so and explain what matters instead, such as clustering or confounding. For small samples, a power analysis with a stated effect size and its assumptions is usually expected. Match the argument to the study.
Which software should be named?
Name what you intend to use, such as R, SPSS, Stata or NVivo, with the version if your instructions ask. The choice rarely affects the grade, but naming it shows the plan is concrete. Where a specific package implements the analysis, such as a mixed-model library, naming it helps a reader judge whether the procedure fits.
Where can I find a free GB730 Unit 9 sample paper?
This page holds a full GB730 Unit 9 data analysis plan for a hotel override study, with mixed models for occupancy and rate, E-values, reflexive thematic analysis and a joint display.
What does an E-value tell a reader?
It gives the minimum strength of association an unmeasured factor would need with both the predictor and the outcome to fully explain away an observed association. A larger E-value means the finding is harder to dismiss.