# Data-quality review: `raw.Social_media_impact_on_life` The table is structurally clean, but it has a few quality problems. Most are missing values, capped or floored values, and one logical inconsistency in age versus academic level. I found no format or unit problems. I only checked ranges, nulls, and boundary counts, not full distributions, so subtle outliers inside the ranges could remain. ## Clean areas - **Keys:** all 4,500 rows have a distinct, non-null `Student_ID`. Every ID is 11 characters long and runs from `STU_2024000` to `STU_2028499`, so there are no duplicates and no odd formats. - **Categorical labels:** the values for gender, level, platform, device, comparison frequency, impact, and late-night usage are consistent, with no case or spelling variants and no nulls. - **Units:** nothing suggests mixed units. - Usage and sleep are in hours. - Sleep quality runs 1–5. - Mental health index runs 32–98. - Stress runs 0–40. - Weekend extra hours run 0–4.5. - **Time budget:** no student has daily usage plus sleep above 20 hours (0 rows), so there are no physically impossible time allocations. - **Age and level:** no postgraduates are aged 20 or under, and no undergraduates are aged 16 or under. ## Issues found 1. **Missing values.** - `Academic_Performance_GPA` is null in 85 rows (1.9%). - `Perceived_Stress_Score` is null in 46 rows (1.0%). - Every other column is fully populated. 2. **Age inconsistent with academic level.** 156 of the 1,477 High School students (about 10.6%) are aged 19 or older. That is implausible for high school and suggests mislabeled level or age. 3. **Suspicious zero stress scores.** 87 rows have a stress score of exactly 0.0, about 2% of the 4,454 non-null scores. A true zero on a perceived-stress scale is rare, so this may be a placeholder or default value. Only 8 rows sit at the maximum of 40. 4. **GPA ceiling pile-up.** 353 rows have a GPA of exactly 4.0, about 8% of non-null GPAs. This looks like capping or clipping, which would compress variation among high achievers. The minimum is 1.9. 5. **Capped or floored behavioural values.** - 55 students sit exactly at the 14.0 daily-usage maximum, which suggests a cap. - 126 students report 12 or more hours a day, which is extreme but not impossible. - Sleep has 11 rows at the 3.0-hour floor and 1 row at the 10.5-hour maximum. 6. **Skewed categories.** These are not errors, but they limit subgroup analysis. - Postgraduates number 246. - Tablet users number 129. - "Negative" impact has 165 rows against 3,681 "Beneficial". - Non-Binary has 122 rows, and "Prefer not to say" has 93. 7. **Metadata mislabel.** The card lists `Daily_Usage_Hours` and `Weekend_Extra_Hours` as temporal axes, but they are durations, not dates. The only true timestamp is `_ingestion_timestamp`, which is a pipeline field. ## Suggested handling - Treat the null GPA and stress rows explicitly, either by excluding them or by flagging them as missing. - Check whether the 87 zero-stress rows are real values or defaults. - Review the 156 High School rows aged 19 or older. - Be cautious with averages and correlations that involve GPA, stress, or usage, because of the capped values. - Avoid drawing strong conclusions from the small subgroups.