PHLT 8032 Week 3 Data Cleaning Log Example

Course	PHLT 8032 SPSS Revealed
Week	Week 3
Paper type	Cleaning log
Length	About 1,200 words, 7 pages
Format	APA 7 student paper
School	Walden University
Program	PhD in Public Health
Updated	October 2026

Free sample paper for PHLT 8032 Week 3

Cleaning Log for a Composite County Health Survey File: Dated Entries for Special Codes, a Duplicate Interview, Implausible Body Measures and Activity Outliers, and How the Working Sample Changed

Student Name

Walden University

PHLT 8032: SPSS Revealed

Instructor Name

Month Day, Year

What this page is doingThe title names each class of problem the log handles, which lets a reader find the decision behind any later figure.

Cleaning Log for a Composite County Health Survey File: Dated Entries for Special Codes, a Duplicate Interview, Implausible Body Measures and Activity Outliers, and How the Working Sample Changed

How This Log Works

This log records every change made to the composite county extract after setup, in the order the changes were made. Each entry gives the date, the variable, what was found, what was done, the reason, the syntax that did it and the number of cases touched. No original variable was overwritten except to correct a keying error; recodes and exclusions were written into new variables with a suffix, so the source values remain available if a decision is questioned. Cleaning has been described as a cycle of screening, diagnosis and editing, and editing without diagnosis produces a tidy file built on guesses (Huebner et al., 2016). Each entry below therefore states the diagnosis before the edit. The definitions from last week fix what counts as a plausible value for each variable.

What this page is doingStating the rule that originals are never overwritten tells the grader every edit can be reversed and checked.

The Log

Table 1

Cleaning Log, Entries 1 to 6

Entry and date	Variable	Found	Done	Reason	Syntax	Cases \|
1, Sept 15	ment_days	Code 88 (none) present	Recoded 88 to 0 into ment_d	88 is a true zero in the codebook	RECODE ment_days (88=0) (ELSE=COPY) INTO ment_d.	1,032 \|
2, Sept 15	phys_days	Code 88 (none) present	Recoded 88 to 0 into phys_d	Same codebook rule	RECODE phys_days (88=0) (ELSE=COPY) INTO phys_d.	1,104 \|
3, Sept 15	ment_d, phys_d	Codes 77 and 99	Declared user-missing on new variables	No information on days	MISSING VALUES ment_d phys_d (77, 99).	30 and 26 \|
4, Sept 16	resp_id	ID 011392 on two rows	Second row deleted	Identical on all 16 fields; interviewer log shows one call	SELECT IF NOT (dup_flag = 1).	1 \|
5, Sept 16	height_in	Values 5, 6 and 6	Set missing in height_c	Feet entered; inches cannot be recovered	IF (height_in LT 48) height_c = $SYSMIS.	3 \|
6, Sept 16	height_in	Values 165 and 178	Converted to inches in height_c	Plausible centimeters; paired weights plausible	IF (height_in GT 96) height_c = height_in / 2.54.	2 \|

Syntax lines are abbreviated where a preceding COMPUTE created the new variable; the full file is kept with the log.

Table 2

Cleaning Log, Entries 7 to 12

Entry and date	Variable	Found	Done	Reason	Syntax	Cases \|
7, Sept 17	weight_lb	Value 1,850 with height 70	Set missing in weight_c	Likely decimal slip, but unverifiable	IF (weight_lb GT 700) weight_c = $SYSMIS.	1 \|
8, Sept 17	bmi	Computed from cleaned measures	New variable, range 14.9 to 61.3	Needed for the definition week's category	COMPUTE bmi = 703 * weight_c / (height_c ** 2).	1,662 valid \|
9, Sept 17	bmi	Four values under 16	Kept after review	Height and weight each plausible for a small adult	None (reviewed with LIST)	4 \|
10, Sept 18	act_min	Values of 4,900, 6,720 and 8,400	Set missing in act_c	Over 10 hours of activity every day	IF act_min GT 4200 act_c = $SYSMIS.	3 \|
11, Sept 18	inc_cat	Code 9 (refused)	Confirmed user-missing; pattern checked	Refusal is not an income	CROSSTABS inc_miss BY age_grp.	164 \|
12, Sept 18	srh, insured	Codes 7 and 9	Confirmed user-missing	Not answers to the question	FREQUENCIES srh insured.	6 and 10 \|

Counts refer to the 1,846 imported rows for entries 1 to 4 and to the 1,845 retained rows afterward.

What this page is doingWriting the syntax beside each decision lets a reviewer rerun any single step without reading the rest of the file.

The Decisions That Changed the Analytic Sample

The Special Codes

Entries 1 to 3 had the largest effect on any later figure. Before the recode, the highest value of mental health days was 88 and its mean was meaningless; after it, 1,116 of the 1,815 respondents with a valid answer report zero poor mental health days, about 61.5%. That figure includes 84 respondents who answered "0" directly instead of choosing the "none" option, which is why it exceeds the count of 88 codes. Declaring 77 and 99 missing removed 30 respondents from mental health analyses and 26 from physical health analyses. They are kept in the file and counted in every report.

The Duplicate

Entry 4 removed one row. The two rows for identifier 011392 matched on every variable, and the interviewer log showed a single completed call. Keying the same interview twice is a data-entry fault rather than a second respondent, so deletion was a correction rather than an exclusion. The working file holds 1,845 respondents.

Body Measures

Entries 5 to 9 concern height and weight, which are self-reported in this survey. Self-reported height tends to be overstated and weight understated, so even plausible values carry some error (Connor Gorber et al., 2007). That error was accepted; only values that could not be a human measurement were edited. Three heights of 5 or 6 were almost certainly feet, but the inches were never recorded, so guessing 5 feet 6 inches would have been invention. They were set missing. Two heights of 165 and 178 fit centimeters exactly and were paired with weights that gave ordinary BMIs once converted, so they were corrected. One weight of 1,850 pounds was paired with a height of 70 inches; a slipped decimal is likely but not provable, so it was set missing. Four BMIs under 16 were reviewed case by case. Each came from a short, light respondent whose height and weight were individually plausible, and they were kept. In all, 1,662 respondents have a valid BMI; most of the 183 without one refused height or weight at interview.

Activity Minutes

Entry 10 set three activity values missing. A weekly total of 8,400 minutes is 20 hours a day. The rule chosen was 4,200 minutes, an average of 10 hours a day, which is generous for physically demanding jobs and excludes only values that cannot describe a week. The rule was set before looking at which respondents it would affect, and the log says so. After it, 1,790 respondents have valid activity minutes.

What this page is doingGiving the cutoff and saying it was chosen before the cases were inspected answers the question a skeptical grader will ask about outliers.

Missing Data That Was Not Edited

Income was refused by 164 respondents, 8.9% of the working file, the largest gap among the analysis variables. Entry 11 checked whether refusal was spread evenly and found it was not: 12.4% of respondents aged 65 or older refused income, compared with 7.6% of younger respondents. Analyses involving income will therefore describe older adults somewhat less fully. Sterne et al. (2009) note that complete-case analysis is unbiased only under restrictive conditions and that the pattern of missingness should be reported whenever it is used. This course does not attempt imputation, so the log records the pattern and later write-ups will state it.

Working Sample Counts for Planned Analyses

The working file has 1,845 respondents. Complete-case counts differ by analysis because each draws on different variables: 1,805 for poor mental health days by insurance status, 1,829 for fair or poor health by insurance status, and 1,675 for self-rated health by income bracket. Pallant (2020) recommends reporting the number of cases behind every result, and these three counts will be checked against the Case Processing Summary each time a procedure runs. A difference between a printed count and these figures will mean that a filter or exclusion was applied without a log entry, which is the first thing to investigate.

References

Connor Gorber, S., Tremblay, M., Moher, D., & Gorber, B. (2007). A comparison of direct vs. self-report measures for assessing height, weight and body mass index: A systematic review. Obesity Reviews, 8(4), 307-326. https://doi.org/10.1111/j.1467-789X.2007.00347.x

Huebner, M., Vach, W., & le Cessie, S. (2016). A systematic approach to initial data analysis is good research practice. The Journal of Thoracic and Cardiovascular Surgery, 151(1), 25-27. https://doi.org/10.1016/j.jtcvs.2015.09.085

Pallant, J. (2020). SPSS survival manual: A step by step guide to data analysis using IBM SPSS (7th ed.). Routledge.

Sterne, J. A. C., White, I. R., Carlin, J. B., Spratt, M., Royston, P., Kenward, M. G., Wood, A. M., & Carpenter, J. R. (2009). Multiple imputation for missing data in epidemiological and clinical research: Potential and pitfalls. BMJ, 338, Article b2393. https://doi.org/10.1136/bmj.b2393

PHLT 8032 Week 3 instructions, in plain terms

The third assignment usually asks for a cleaning log covering every change made to the course data after setup. Faculty want each decision traceable: a date, the variable, the problem found, the action taken, the reason and the syntax that did it, with a count of affected cases. Logs that recode into new variables, leaving the originals untouched, are preferred because a questioned decision can be reversed. Outlier rules should be stated as rules and set before looking at which respondents they remove. Many prompts also ask for a short narrative after the table explaining which decisions changed who is analyzed, how missing data are distributed, and how many cases each planned analysis will use. Keeping routine checks brief while giving judgment calls a full paragraph shows the student knows which decisions matter.

How the PHLT 8032 Week 3 example is put together

A short introduction sets the rule that originals are never overwritten except to fix a keying error and frames the log with the screening, diagnosis and editing stages from the data cleaning literature. Two log tables follow with twelve entries in all, each carrying its syntax line and case count. The narrative takes the special codes first, showing that 61.5% of valid respondents report no poor mental health days once 88 is recoded. It explains why the duplicate removal is a correction, why heights in feet were set missing rather than guessed, and why four very low BMIs were kept after review. The activity cutoff is defended as generous. A section on income refusal reports its uneven spread by age, and the log closes with complete-case counts for each planned analysis.

Reading the PHLT 8032 Week 3 rubric

Graders reward traceability above all. An entry missing its case count or its syntax line cannot be checked, and an unexplained exclusion is the most common reason a log loses marks. The special-code entries are read closely, since recoding 88 to zero and declaring 77 and 99 missing changes every later figure. Outlier handling earns credit when the cutoff is stated, justified and fixed in advance, and loses it when values are dropped because they look odd. Credit also goes to keeping originals, diagnosing before editing, and saying when a likely correction could not be proven. The narrative should report the pattern of missing data rather than only its amount, and the closing counts should match what each later procedure prints. Sources on cleaning and missing data support the judgment calls.

PHLT 8032 Week 3 help: mistakes that cost marks

The weakest logs list what was done without why, or describe decisions so vaguely that nobody could repeat them. Write each entry so a stranger with your syntax and the raw file would reach the same result. Recode into new variables. When a value is impossible, say what makes it impossible and whether the true value can be recovered; if it cannot, set it missing instead of guessing. Fix outlier rules before you look at who they affect, and say that you did. Report how missing answers are spread across groups, not only how many there are, because a gap concentrated among older respondents changes what later results describe. Finish with the case count each analysis should show. A log written against your own section's file and codebook can be built in the same format.

Get a PHLT 8032 Week 3 example written to your instructions

Share the Week 3 prompt and rubric, and describe in general terms what went wrong in setup. The log comes back entry by entry with rules, reasons and syntax references, in 24-48h, with your first request free. Case counts in it are placeholders; those in your log must come from your file.

More PHLT 8032 papers and PhD in Public Health samples

PHLT 8032 Week 3 questions, answered

Is it acceptable to delete cases with missing values?

Sometimes, but the log has to say how many, which ones and why, and what the deletion does to the sample. Listwise exclusion is the default in many procedures, and it quietly drops anyone missing any variable in the analysis. Stronger logs report the complete-case count for each planned analysis and name the groups most affected, rather than letting the software decide silently.

Should recodes overwrite the original variable?

Rarely. Recoding into a new variable keeps the original available if the decision is later questioned, and it lets a reader compare the two. Most rubrics reward that practice, and the log records both names. Overwriting is defensible only for pure corrections, such as fixing a keying error, and even then the entry states the original value so the change can be reversed.

How detailed does each log entry need to be?

Enough that a stranger to the file could repeat the decision without contacting you. That usually means the variable, the rule applied, the number of cases affected and the syntax line. A sentence of reasoning covers routine decisions; judgment calls, such as excluding a value with no traceable cause, deserve a short paragraph acknowledging the alternative that was rejected.

Where can I find a free PHLT 8032 Week 3 sample paper?

This page holds the complete Week 3 cleaning log, with twelve dated entries, the SPSS syntax for each, and a narrative on how the working sample of 1,845 respondents was reached.

Is setting an impossible height to missing better than correcting it?

Only when the true value cannot be recovered. Heights in centimeters convert exactly, so they were corrected; heights entered in feet carry no inches, so a correction would have been a guess.