Lost uploads, broken-off interviews, duplicates and impossible hours are each screened, diagnosed and treated in PHLT 8068's data cleaning memo for Week 5, with every change logged. Searches like "phlt 8068 week 5 assignment example", "phlt8068 week 5 sample" and "phlt 8068 week 5 example" land here.
What a finished PHLT 8068 Week 5 data cleaning memo looks like
A six-page memo with an appendix log, it takes each class of problem in turn under three headings borrowed from Van den Broeck and colleagues: how it was screened, what diagnosis was reached, and what treatment followed. Tablet interviews that failed to upload are counted against the interviewers' paper tally sheets, and the memo says which were recovered from device backups and which were lost. Partial interviews are kept where the symptom section was complete. Duplicates are found through a screening question and a self-generated code, and the later interview is dropped. Implausible values, such as daily work hours beyond what a day allows, are set to missing rather than trimmed. A section on item missingness notes that questions about employer-provided rest had the most blanks. The appendix lists every change with record identifier, rule and date.
How a PHLT 8068 Week 5 example is structured
Organizing by problem rather than by variable lets each rule be judged once and applied everywhere, which is what makes cleaning reproducible. The three headings keep diagnosis separate from treatment: deciding that an hours value is impossible is a different judgment from deciding what to do about it, and merging them hides the second choice. Lost uploads come first because they affect the denominator the recruitment report already published, and the memo reconciles the two. Rules were written before outcome associations were examined, and the memo says so, since cleaning decisions made after seeing results invite suspicion. The missingness section brings Rubin's classification of missing data to the employer questions, asking whether their blanks could depend on the unrecorded answer itself. The dated log is the memo's evidence and sits in the appendix.
Screen, diagnose, treat
Each problem is described three times: how it surfaced, what was concluded about it, and what was done.
Uploads that failed
Tablet interviews that never synced are counted against paper tally sheets. Some came back from device backups; the rest are reported as lost.
Twice interviewed
A screening question and a self-generated code flag repeat interviews across sites. The later one is removed.
Values set to missing
Work hours no day could hold are treated as missing, not trimmed to a ceiling, so no value is invented.
Blanks on employer questions
Questions about rest breaks drew the most missing answers. The memo considers whether the blanks depend on the answers themselves.
Where marks go in PHLT 8068 Week 5
Traceability is the standard. A memo announcing that cleaning happened and outliers went describes an outcome without a process, and faculty read that as a gap in the chain between collection and results. The dated log and the separation of diagnosis from treatment carry this example, since together they let another analyst repeat every decision. Reconciling lost uploads with the recruitment report is weighted because it keeps the study's denominators consistent. Setting impossible values to missing, rather than trimming, is credited as a choice that avoids inventing data. The missingness section is rewarded for applying Rubin's categories to a specific pattern instead of reciting definitions. Memos lose standing when rules appear to have been written after results were seen, when dropped records go uncounted, or when imputation appears with no word on which variables it touched.
Get a PHLT 8068 Week 5 example written to your instructions
Share your raw data dictionary, the problems you have found so far and the Week 5 memo instructions with the rubric. Each problem is screened, diagnosed and treated under stated rules, with a dated log of every record changed. First memo free, back within 24 to 48 hours. The survey records it cleans were made up for the page.
PHLT 8068 Week 5 questions, answered
Who are Van den Broeck and colleagues?
Authors of a widely cited paper describing data cleaning as a process with distinct phases: screening for suspect data, diagnosing whether each is an error, a true extreme or undetermined, and treating it by correcting, deleting or leaving it unchanged. They stress documenting every step. The example borrows that structure for each class of problem, which keeps judgment about a value separate from the action taken.
Should outliers be deleted?
Only values that are impossible, and even then setting them to missing is usually better than deleting the whole record. A true extreme, such as a very long shift during harvest, is data and should stay. The example sets impossible hours to missing and keeps long but possible ones. Your rules should say which is which before you look at how any value relates to outcomes.
What if missing answers cluster on sensitive questions?
Then the pattern itself is a finding worth reporting. The example's blanks gathered on employer-related rest questions, which suggests missingness may depend on the unobserved answer, the case Rubin called missing not at random. No method fully repairs that. Report the pattern, run analyses under different assumptions, and carry the concern into your limitations rather than burying it in the cleaning.