| Course | GB701 Business Statistics |
|---|---|
| Unit | Unit 4 |
| Paper type | Sampling and power memo |
| Length | About 810 words, 5 pages double-spaced |
| Format | APA 7 student paper |
| School | Purdue University Global |
| Program | MBA |
| Updated | October 2026 |
Free sample paper for GB701 Unit 4
How Many Files Before Opening Any: A Unit 4 Sampling and Power Memo for a Quote Conversion Study
[Student Name]
Purdue University Global
GB701: Business Statistics
Unit 4 Sampling and Power Memo
[Instructor Name]
[Date]
The distributor, its accounts and every figure are composite and illustrative; the software, methods and sources cited are real.
Memo
TO: Doctoral mentor FROM: DBA candidate, director of sales operations SUBJECT: Sample size and design for the quote conversion study
The Question the Data Must Answer
The study asks whether accounts that request quotes through the customer portal convert quotes to orders faster than accounts that request quotes by email or phone, once order type, the number of heavily revised quotes, account tenure and annual purchases are held constant. Portal use cannot be read from the quoting system, because portal requests are logged in a separate system that records the user's login but not the account reliably. Revision rounds must be counted from each quote's history. Both require opening each account's records by hand, about forty minutes per account, so coding all 1,240 accounts would take more than 800 hours.
Two Analyses, Two Power Runs
Primary analysis. A multiple regression of quote-to-order time on five predictors, testing whether portal use adds explained variance once the other four are included. In G*Power 3.1 (Faul et al., 2009), the test family is F tests, linear multiple regression, fixed model, R-squared increase, run a priori. With an effect size f-squared of .04, between the small and medium benchmarks Cohen (1988) proposed, alpha of .05, power of .80, one tested predictor and five predictors in total, the required total sample is 199.
Secondary analysis. An unadjusted comparison of mean quote-to-order time between portal users and other accounts, using Welch's t test. In G*Power, the family is t tests, means, difference between two independent means, two groups, a priori. The smallest difference worth detecting is two days, explained below. With the population standard deviation of 5.01 days from the full quoting-system extract, the effect size d is 0.40. Portal users are expected to be about a third of accounts, so the allocation ratio is set at 1.9 non-users per user. With alpha of .05 and power of .80, two-tailed, the required sizes are 76 portal users and 144 others, a total of 220.
The larger requirement, 220, governs the plan.
Where Two Days Came From
The vice president of sales was asked, before any data were examined, what difference in conversion time would change a decision about promoting the portal to customers. Her answer was two days: a portal that sped conversion by less would not justify the cost of moving customers off email, while two days would shorten the cash cycle on about $460 million of annual sales enough to matter. A smallest effect of interest set by the decision-maker, rather than by convention or by what the data later show, is the defensible way to choose it.
Allowance for Files That Cannot Be Coded
Some accounts will not be codable. A pilot of 20 accounts found one with a portal login shared by several branches of the customer, which makes portal use impossible to attribute, and one with an incomplete quote history after a system migration. Allowing for about 6 percent of files to fail, the plan draws 234 accounts, since 220 divided by 0.94 is 234.
Stratified by Order Type, Allocated Proportionally
Unit 3 showed that slow conversion is concentrated among special-order accounts, which are 28.1 percent of the frame. The sample is therefore stratified by order type with proportional allocation: 66 special-order accounts and 168 stock-order accounts, drawn at random within each stratum using a seeded random number generator whose seed is recorded in the analysis log. Proportional allocation keeps the sample self-weighting, so means and regression estimates need no sampling weights. Lohr (2019) notes that stratification with proportional allocation can only reduce the variance of a mean compared with simple random sampling when strata differ, as these do.
The sample size was fixed by the smallest difference the business cares about, not by how much coding time was left.
What the Memo Declines to Promise
The memo does not promise that a difference will be found, only that a two-day difference would be detected with 80 percent probability if it exists. It does not promise power for subgroup analyses, such as portal effects within special-order accounts alone; those would be exploratory. And because the data are archival, the study can estimate an association between portal use and conversion time, not the effect of moving a customer onto the portal, since customers chose whether to use it.
Next Steps
The 234 accounts will be drawn this week, coding will follow over six weeks, and a second coder will independently code a random 10 percent of files to check agreement on portal attribution and revision counts.
References
Cohen, J. (1988). Statistical power analysis for the behavioral sciences (2nd ed.). Lawrence Erlbaum Associates.
Faul, F., Erdfelder, E., Buchner, A., & Lang, A.-G. (2009). Statistical power analyses using G*Power 3.1: Tests for correlation and regression analyses. Behavior Research Methods, 41(4), 1149-1160. https://doi.org/10.3758/BRM.41.4.1149
Lohr, S. L. (2019). Sampling: Design and analysis (2nd ed.). CRC Press.
Reading the GB 701 Unit 4 instructions
The fourth GB701 assignment asks you to decide how much data your question needs before collecting it, in a memo to your mentor. State the question and the analyses that will answer it. For each analysis, report a power calculation with the software, test family, effect size, alpha, power and any allocation ratio, and let the largest requirement govern. Justify the effect size, ideally from a decision-maker's threshold rather than convention alone. Add an allowance for unusable cases based on evidence such as a pilot. Describe the sampling design, including any stratification and how random selection will be documented. State what the plan cannot promise. Close with next steps. About three pages plus appendix screens, cited in APA 7, is common.
How this GB701 Unit 4 example is built
The memo explains why hand coding forces sampling. The primary analysis, a regression R-squared increase test in G*Power with f-squared of .04, returns 199, citing Faul and colleagues and the benchmarks Cohen proposed. The secondary Welch comparison, d of 0.40 with a 1.9 allocation ratio, returns 76 and 144. A section explains where the two days came from. A pilot of 20 accounts supports a 6 percent allowance, giving 234. Proportional stratification yields 66 special-order and 168 stock accounts with a recorded seed, citing Lohr, and the highlighted line ties the size to the business threshold. A section lists what the memo will not promise, and second-coder checks close it.
Where the points sit in the GB701 Unit 4 rubric
GB701 instructors grade the sampling memo on whether every number can be traced. Power calculations reported with software, test family and all inputs earn heavy weight because another researcher could reproduce them. Running separate calculations for each planned analysis, and letting the largest govern, shows the sample serves the whole study. An effect size justified by a decision threshold set before the data shows rigor. An allowance for unusable cases based on a pilot demonstrates realism. A documented random selection and a reasoned stratification design show competence. Plain limits on what the sample can deliver, and APA 7 references for G*Power and the power conventions, close out the GB701 memo.
GB701 Unit 4 help: mistakes that cost points
The most common GB701 power memo error is reporting a sample size without the inputs that produced it, so no one can check it. Another is choosing an effect size because it gives a convenient number. Many memos run a power analysis for the main test only, leaving secondary analyses underpowered. Writers often forget that some files will be unusable. Sampling designs are sometimes described as random without saying how selection will be done or recorded. Report every input, run power for each analysis, justify the effect size from the decision, add a pilot-based allowance, document the random draw and state what the sample cannot do. Request a free custom sampling and power memo for your study.
Get a GB701 Unit 4 example written to your instructions
Tell us the analysis your study plans, the variables involved and any effect size your section or prior research suggests, and include the Unit 4 prompt with its rubric. Laid out for checking, the G*Power inputs and outputs arrive in a custom memo within 24-48h. Your first one is free. 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 GB701 papers and MBA samples
- GB701 Unit 1: Discussion Board Post
- GB701 Unit 2: Descriptive Summary Exercise
- GB701 Unit 3: Probability Problem Set
- GB701 Unit 5: Confidence Interval Exercise
- GB701 Unit 6: Seminar Reflection
- GB701 Unit 7: Hypothesis Test Writeup
- GB701 Unit 8: Regression Model Report
- GB701 Unit 9: Assumption Diagnostics Memo
- GB701 Unit 10: Results Section Draft
- GB700 Unit 4: Source Evaluation Memo
- GB601 Unit 4: Competitive Positioning Evidence
- GB600 Unit 4: Problem Classification Exercise
- GB590 Unit 4: Governance Failure Case
GB701 Unit 4 questions, answered
Which G*Power test fits a single predictor in a regression?
For testing whether one predictor adds explained variance beyond the others, the F tests family offers Linear multiple regression: Fixed model, R-squared increase, with the number of tested predictors set to one and the total number of predictors entered separately. The R-squared deviation from zero test asks a different question, whether the model as a whole explains anything, and can return a very different sample size.
What if the available data fall short of the power analysis?
Say so and report a sensitivity analysis, which gives the smallest effect the available sample can detect with adequate power. Committees generally prefer an honest sensitivity statement to a power analysis reverse-engineered to match the data on hand. The limitation then belongs in the discussion of any null result, where it matters most to a reader.
Is 80 percent power always the target?
It is a convention, not a rule. Eighty percent accepts a one-in-five chance of missing a real effect of the planned size. Where a missed effect would be costly, 90 percent may be justified, at the price of a larger sample. Name the choice and its consequence, since instructors usually mark the reasoning rather than the figure itself.
Where can I find a free GB701 Unit 4 sample paper?
The complete GB701 Unit 4 sampling and power memo on this page reaches 234 accounts through two G*Power analyses, a decision-based effect size, a pilot-based allowance and proportional stratification.
Why let the larger power requirement govern?
Because the sample must support every planned analysis. Using the smaller figure would leave the secondary test underpowered, so a result there could be missed for lack of data.