# Distribution summary of the numeric columns (300 rows in the table) Each column is summarized over its valid (parseable) values only. Non-numeric or missing entries were excluded, so `n` varies by column. | Column | Valid n | Missing / unparsed | Min | Median | Mean | Max | Std dev | |---|---|---|---|---|---|---|---| | Age | 263 | 37 | 16.0 | 27.0 | 30.18 | 97.0 | 12.47 | | Monthly_Income | 290* | 10 | 0.0 | 521.5 | 2,304.38 | 250,000 | 14,743.44 | | AI_Usage_Hours_Per_Day | 271 | 29 | 0.5 | 2.2 | 3.01 | 23.0 | 2.45 | | Monthly_AI_Cost | 288* | 12 | 0.0 | 0.0 | 9.84 | 100.0 | 13.75 | | Time_Saved_Hours_Per_Week | 272 | 28 | 0.0 | 5.05 | 5.81 | 19.4 | 4.11 | | Work_or_Study_Hours_Per_Day | 279 | 21 | 2.1 | 7.2 | 7.29 | 20.0 | 2.14 | | Productivity_Score | 284 | 16 | 1.6 | 5.0 | 5.23 | 10.0 | 1.82 | | Accuracy_Rating | 285 | 15 | 2.0 | 4.0 | 3.43 | 5.0 | 0.86 | | Satisfaction_Score | 293 | 7 | 3.0 | 6.0 | 6.15 | 10.0 | 1.52 | \*See the caveat below on Monthly_Income and Monthly_AI_Cost. ## Spread beyond the standard deviation Quartiles and the 95th percentile (p95) show how each column is shaped: | Column | p25 | p75 | p95 | |---|---|---|---| | Age | 22.0 | 35.0 | 54.9 | | Monthly_Income | 110.75 | 1,627.25 | 6,399 | | AI_Usage_Hours_Per_Day | 1.4 | 4.1 | 7.55 | | Monthly_AI_Cost | 0.0 | 20.0 | 30.0 | | Time_Saved_Hours_Per_Week | 2.7 | 7.75 | 13.54 | | Work_or_Study_Hours_Per_Day | 6.1 | 8.4 | 10.52 | | Productivity_Score | 3.9 | 6.5 | 8.6 | | Accuracy_Rating | 3.0 | 4.0 | 5.0 | | Satisfaction_Score | 5.0 | 7.0 | 9.0 | ## What the data shows - **Monthly_Income is extremely right-skewed.** The mean (≈2,304) is about 4.4 times the median (521.5). The standard deviation (≈14,743) is far larger than the mean. The max of 250,000 is nearly 40 times the p95 of 6,399, so a few extreme values drive the mean and spread. The median and quartiles describe a typical respondent better. - **Monthly_AI_Cost is zero for at least half of respondents.** The median and p25 are both 0, and p75 is 20. The mean of 9.84 is therefore pulled up by a minority who pay, with a max of 100. - **AI_Usage_Hours_Per_Day has a long right tail.** The median is 2.2 and p95 is 7.55, but the max is 23.0 hours. That is implausible as real daily usage and may be a data-entry error. - **Age** is centered on the late 20s (median 27, middle half 22–35). The max of 97 is an outlier; p95 is only 54.9. - **Work_or_Study_Hours_Per_Day** is tightly centered (median 7.2, mean 7.29), with an extreme max of 20.0 against a p95 of 10.52. - **Time_Saved_Hours_Per_Week** is mildly right-skewed (mean 5.81 vs median 5.05). - **Productivity_Score and Satisfaction_Score** are roughly symmetric (mean ≈ median), and nobody scored below 1.6 and 3.0 respectively. - **Accuracy_Rating** is a coarse 2–5 scale with a median of 4 and a slightly lower mean (3.43), so it is mildly left-skewed. ## Caveats and open points - **Two versions of Monthly_Income and Monthly_AI_Cost appeared.** Different query runs parsed different numbers of valid values. - Version A (my primary, shown in the main table): income n=290, mean 2,304.38, median 521.5; cost n=288, mean 9.84. - Version B: income n=284, mean 2,328.96, median 520.0, quartiles 112.25 / 1,617.75; cost n=273, mean 9.76. - The evidence doesn't say which parsing is more correct. Version A accepted more raw text values as numeric, probably formatting differences such as symbols or separators, but I didn't verify this. The conclusions above hold under both versions. - **Missing and unparsed values are substantial.** Age alone has 37 unparsed values (about 12%). For Age, AI usage, time saved and work hours, the excluded values failed numeric conversion. For the three rating/score columns, they are simply null. I did not inspect what the unparsed text looks like, so some could be recoverable. - **Outliers were not removed or verified.** Max values such as Age 97, income 250,000, and 23 hours of daily AI use are reported as found. Whether they are errors or genuine is not established.