| Course | GB892 Doctoral Project III: Data Analysis |
|---|---|
| Unit | Unit 2 |
| Paper type | Data preparation record |
| Length | About 940 words, 6 pages double-spaced |
| Format | APA 7 student paper |
| School | Purdue University Global |
| Program | MBA |
| Updated | October 2026 |
Free sample paper for GB892 Unit 2
From 288,061 Records to 196,518 Visits: A Unit 2 Data Preparation Record for a Dock Appointment Study
[Student Name]
Purdue University Global
GB892: Doctoral Project III: Data Analysis
Unit 2 Data Preparation Record
[Instructor Name]
[Date]
Haverly Logistics, its yard system and every count are composite and illustrative; the methods sources cited are real.
The Source
The extract contains every inbound gate record from Haverly's twelve distribution centers over eighteen months, exported from the yard management system as one file with one row per gate-in event. Each row carries a visit identifier, the center, the carrier, the trailer number, the load type, live-unload or drop-trailer, and four timestamps: gate-in, door assigned, unload complete and gate-out. The raw file is stored read-only, with a checksum recorded on the day of export, and is never edited. Every change is made by a script that reads the raw file and writes a new one, so the path from raw data to analysis file can be rerun and checked. Wickham (2014) describes this separation between raw data and the steps that tidy it as essential to reproducible analysis.
The Exclusion Rules, in Order
The order matters, because some records fail more than one rule, and the count for each rule depends on what earlier rules removed. The rules are applied as follows.
Rule 1: Drop-trailer visits. 84,823 records. In a drop-trailer visit, a carrier leaves a loaded trailer in the yard and departs, and the trailer is unloaded later. Dwell time for these visits measures how long the trailer sat, not how long a driver waited, and appointments apply only to live unloads. These visits are outside the study by design. Script lines 41 to 44.
Rule 2: Duplicate gate-in scans. 1,817 records. A duplicate is defined as the same trailer number at the same center with a gate-in time within ten minutes of an earlier gate-in. Duplicates arise when a guard scans a trailer twice. The earlier record is kept. Script lines 47 to 55.
Rule 3: Missing gate-out. 2,896 records. Without a gate-out time, dwell cannot be calculated. Most of these records belong to trucks that left without being scanned out. Script lines 58 to 60.
Rule 4: Departures auto-filled at 23:59. 1,291 records. During one center's yard system migration in month seven, the old system filled any missing gate-out with 23:59 on the day of arrival. These times are not real departures. The rule removes records from that center in that month with a gate-out of exactly 23:59. Script lines 63 to 69.
Rule 5: Implausible dwell. 716 records. Dwell under ten minutes cannot include unloading, and dwell over twenty-four hours almost always reflects a missed scan. Both are removed. Script lines 72 to 75.
After the five rules, 196,518 live-unload visits remain.
The Flow
The flow of records runs from 288,061 raw records to 203,238 after removing drop-trailer visits, 201,421 after removing duplicates, 198,525 after removing missing gate-outs, 197,234 after removing the auto-filled times and 196,518 after removing implausible dwell.
Are the Dropped Visits Different?
Exclusions can bias results if the dropped records differ systematically between adopting and comparison centers, or before and after adoption, and with only twelve centers supplying the comparison, inference is already fragile (Bertrand et al., 2004). The record therefore compares the 6,720 live-unload records removed by rules 2 through 5 with the 196,518 retained, on center group, period, load type and time of day.
Excluded records were split between adopting and comparison centers in nearly the same proportion as retained records, 64 percent and 36 percent against 63 and 37. They were somewhat more common in the pre-adoption period, mainly because the auto-filled departures from the migration fell before month ten. They were more often night arrivals, 31 percent against 24 percent of retained visits, consistent with missed scans being more frequent on night shifts.
Night visits were dropped more often than day visits, so any effect concentrated at night will be slightly underrepresented, and the record says so.
Change Log
The change log records every revision to the cleaning script with its date and reason. Two entries matter. The ten-minute duplicate window was originally five minutes; it was widened after inspection showed many duplicate scans six to nine minutes apart at two centers. And the 23:59 rule was added after the comparison of dwell distributions by center showed a spike at exactly that time at one center during its migration.
Checks Run Before the File Was Released
Before the analysis file was released for modeling, three checks were run and their results recorded. The number of visits per center per month was plotted, and no center showed an unexplained gap or spike once the migration weeks were handled. The four timestamps were checked for order, and no retained visit had a door assignment before gate-in or a gate-out before unload completion. And the share of visits by load type and shift was compared across the eighteen months to confirm that the extract had not silently changed what it recorded partway through.
What the Record Allows
Because the raw file is untouched and every rule is scripted, counted and logged, a reader can rerun the preparation, change any rule and see how many records and which results it affects. The analysis in later units will report a sensitivity check in which rule 5's limits are widened, so that the effect of the implausible dwell rule on the estimate is visible. Angrist and Pischke (2009) stress that robustness to reasonable alternative choices is part of what makes an estimate credible, and the record makes those alternatives easy to run.
References
Angrist, J. D., & Pischke, J.-S. (2009). Mostly harmless econometrics: An empiricist's companion. Princeton University Press.
Bertrand, M., Duflo, E., & Mullainathan, S. (2004). How much should we trust differences-in-differences estimates? The Quarterly Journal of Economics, 119(1), 249-275. https://doi.org/10.1162/003355304772839588
Wickham, H. (2014). Tidy data. Journal of Statistical Software, 59(10), 1-23. https://doi.org/10.18637/jss.v059.i10
Reading the GB 892 Unit 2 instructions
Unit 2 of GB892 asks you to record the path from the raw export to the file you will analyze. Describe the source, how it was exported and how the raw file is protected from editing. List each exclusion or cleaning rule in the order applied, with its exact wording, the script lines that implement it and the number of records it removed, explaining why order matters. Show the flow of records from raw count to final count. Then compare excluded and retained records on the variables that define your comparison, such as group and period, and say what any difference implies. Keep a change log of revisions to the rules with reasons. Record any checks run before releasing the file. Close with what the record allows a reader to do, such as rerunning with a different rule. Six pages with an exclusion table and a flow figure is common.
How the GB 892 Unit 2 example is put together
The record describes a one-row-per-gate-event extract stored read-only with a checksum, and cites Wickham on separating raw data from tidying steps. Five rules follow in order, each with its count and script lines: drop-trailer loads, duplicate scans within ten minutes, missing gate-outs, 23:59 departures from one center's migration month and dwell under ten minutes or over a day. A margin note explains the script references. The flow runs to 196,518. Excluded records match retained ones by group but are more often night arrivals, which the highlighted sentence acknowledges, with Bertrand and colleagues on fragile inference. A change log, pre-release checks and a planned sensitivity run, supported by Angrist and Pischke, close the record.
Where the points sit in the GB892 Unit 2 rubric
Points in the GB892 data preparation record go to reproducibility and transparency. Instructors give credit for a protected raw file and scripted changes. Each rule should be stated exactly, in order, with its count and code location, and an explanation of why order matters is valued. A flow from raw to final counts that adds up is expected. Comparing excluded and retained records on the variables central to the comparison is weighted heavily, since exclusions can bias results. Stating what any imbalance implies shows judgment. A dated change log shows how rules evolved. Pre-release checks show care. Explaining how a reader could rerun or vary the preparation demonstrates the record's purpose, and methods sources should be cited.
GB892 Unit 2 help: mistakes that cost points
Many GB892 preparation records describe cleaning in a paragraph without counts, order or code references, so no one can check them. Others edit the raw file directly, leaving no path back. Students often report how many records were dropped but not whether the dropped records differ from those kept. Changes to rules made during cleaning are rarely logged. Checks before release are skipped. Protect the raw file, script every step, list rules in order with counts and code lines, show the flow, compare excluded and retained records, log rule changes, run and record checks and explain how a reader can rerun the work. Send your cleaning steps, and our team can help you write them up as a record.
Get a GB892 Unit 2 example written to your instructions
Attach your raw record counts and whatever exclusions you have already made, even informally, along with the Unit 2 prompt and rubric. We return a preparation record with each rule written out, applied in a stated order and compared against what stayed. Turnaround runs 24-48h, and there is no charge for the first custom sample. 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 GB892 papers and MBA samples
- GB892 Unit 1: Discussion Board Post
- GB892 Unit 3: Codebook
- GB892 Unit 4: Assumption Check Summary
- GB892 Unit 5: Descriptive Results Table
- GB892 Unit 6: Seminar Reflection
- GB892 Unit 7: Thematic Analysis
- GB892 Unit 8: Inferential Results Write-up
- GB892 Unit 9: Departure Log
- GB892 Unit 10: Results Chapter Draft
- GB891 Unit 2: Consent Language Draft
- GB500 Unit 2: Management Concepts Analysis
- GB890 Unit 2: Problem Evidence Memo
- GB512 Unit 2: Audience Rewrite Exercise
GB892 Unit 2 questions, answered
Does a data preparation record count outlier removal as an exclusion?
Yes, and it needs a rule written before any comparison is run. Here the rule is physical plausibility: under ten minutes or over a day cannot be a live unload. Statistical trimming, such as dropping values beyond three standard deviations, is harder to defend with skewed timing data and tends to draw questions about why that threshold was chosen.
How detailed should the comparison of excluded and retained records be?
Detailed enough to show whether the dropped records differ on variables the analysis uses. Shift, site group and load type are the obvious ones in a dock study. A table of counts and percentages is usually sufficient, and formal tests are optional. What matters is stating any imbalance plainly, such as night arrivals missing more gate-outs, and carrying it forward.
Should the record include the cleaning code?
Many sections welcome it as an appendix, and citing script lines in the exclusion table makes the record checkable. The prose still has to stand alone, since a committee member may never open the code. Your program's rules on data sharing and confidentiality apply, particularly where the source organization restricts what leaves its systems.
Where can I find a free GB892 Unit 2 sample paper?
A complete GB892 Unit 2 data preparation record is on this page, tracing a yard system extract from 288,061 records to 196,518 visits under five numbered rules.
Why does the order of exclusion rules matter?
Some records fail more than one rule, and each rule's count depends on what earlier rules removed. Fixing the order makes the counts reproducible and the flow of records add up.