Notes on Bar Graphs, Histograms, Scatter Plots, Data Collection, Correlation, Distribution Shapes, and Statistical Significance

Bar graphs vs histograms

• Bar graph: bars with gaps between categories; used for discrete categories or grouped data.

• Histogram: no gaps between bars; displays frequency distribution for continuous data; essentially showing how often values occur within bins.

• They are largely interchangeable for conveying the same frequency information, when bins/categories are chosen appropriately.

• Important caution: these charts can be manipulated to mislead the audience.

  • Example given: reliability ratings for trucks (Ford, Toyota, Chevy, Ram).

  • Trick: fine print can say data are based on major repairs after ten years, which skews the appearance of reliability.

  • If Ford shows 96% reliability and Toyota 94% after ten years, a naïve reading may claim Ford is clearly more reliable by 2%; in context, the practical difference may be minor.

  • Lesson: read the fine print and consider what the percentages are actually measuring; be cautious with marketing-driven data presentation.

Scatter plots and the regression-to-the-mean concept

• Scatter plots: show clusters of two variables and indicate the strength and direction of their relationship via the slope.

• Relationship directions:

  • Positive correlation: as one variable increases, the other tends to increase.

  • Negative correlation: as one variable increases, the other tends to decrease.

  • No/weak correlation: difficult to discern a clear positive or negative relationship (near zero correlation).

• Regression toward the mean:

  • When a measurement is extreme on the first occasion, the subsequent measurement tends to be closer to the average.

  • Example: a basketball player averages 30 points per game; in the first game of a season, scores 60 points; by season end, scores are expected to be closer to their average rather than remaining at 60.

• Data collection can be qualitative or quantitative (or both):

  • Qualitative data: descriptive information from interviews, observations, case studies.

  • Quantitative data: numerical data (e.g., Likert scales). Example provided: chipotle question: "Will you come back to Chipotle in the next week?" on a 5-point scale (1 = strongly disagree, 5 = strongly agree).

• Likert scale example highlights:

  • Quantitative data (scale 1–5) is cheap and easy to collect but may lack depth.

  • Qualitative questions (e.g., "What do you like about Chipotle?") provide richer detail but are more labor-intensive.

• Scatter plot examples:

  • Positive trend: two variables move upward together.

  • Negative trend example verbally given: more drugs used correlates with lower grades.

  • No clear trend: data scattered with no obvious pattern.

Correlation concepts and the correlation coefficient

• Correlation coefficient measures the strength and direction of a linear relationship between two variables.

• Value range: $r \,\in\,[-1,1]$

  • r ≈ +1: strong positive linear relationship.

  • r ≈ -1: strong negative linear relationship.

  • r ≈ 0: little to no linear relationship.

• Key interpretation caveat: a higher absolute value of r indicates a stronger linear relationship, not necessarily causation.

• Examples:

  • Positive correlation example: more studying, better grades (both variables increase together).

  • Negative correlation example: more drugs, lower grades (as one increases, the other decreases).

• Reminder: correlation does not imply causation; r tells about linear association, not about cause-and-effect.

Distribution shapes, skewness, and related ideas

• Skewness: describes where the bulk of data lies relative to the tails

  • Positive skew: bulk of data on the left, tail to the right.

  • Negative skew: bulk of data on the right, tail to the left.

• Bell curve (normal distribution): classic reference point for many natural phenomena.

• IQ distribution as discussed in the session:

  • The teacher described a bell-curve-like view: 70% of IQs fall within 15% of the mean (mean = 100), roughly between 85 and 115.

  • Also noted: 90% of IQs fall within 30% of the mean, roughly between 70 and 130.

  • The speaker noted these are approximations in a casual discussion.

• Bimodal distribution: two peaks in the data set.

• Aside/aside anecdote: a humorous mention of Ron McDonald and tying it to the bell curve; included as a classroom anecdote rather than a methodological point.

• Practical takeaway: distribution shape affects interpretation of central tendency and variability.

Central tendency measures

• Central tendency concepts describe typical values in a distribution:

  • Mode: the most frequently occurring value (the peak(s) in the distribution).

  • Mean: the arithmetic average; $\bar{x} = \frac{1}{N}\sum<em>{i=1}^N x</em>i$.

  • Median: the middle value when data are ordered; divides the dataset into two halves.

• The session used a narrative example around test scores::

  • If 66 students take a test, the mode might be the score that occurs most often (e.g., many students scoring 16/16 on a vocab test), and the mean might be around 98–99% if most students perform well.

  • The mode is often higher than the mean when a large portion of students is well-prepared and there are a few low outliers pulling the mean down.

• Median is the middle score, with half above and half below; example discussed with questions of exact placement for a class size (illustrative purposes).

• Range context:

  • Range = max(xi) − min(xi).

  • Example: scores from 6% to 96% yield a range of 90 percentage points.

  • It is common to exclude the highest and lowest scores to get a clearer sense of typical performance.

Spread, variability, and the standard deviation concept

• Range describes the overall spread (from max to min), but it is sensitive to outliers.

• Standard deviation measures how spread out data are around the mean; lower standard deviation indicates more consistency.

• Illustrative basketball example:

  • Player A: scores 50, 10, 12, 52, 14 (high variability).

  • Player B: scores 35, 28, 32, 30, 31 (more consistent).

  • If you’re building a team, you would prefer the player with the lower standard deviation (more reliable scoring).

• Everyday application for test design: a test writer aims for a relatively low standard deviation among test scores, allowing for a predictable distribution with a few outliers.

• Formulas:

  • Population standard deviation: $\sigma = \sqrt{\frac{1}{N}\sum<em>{i=1}^N (x</em>i - \mu)^2}$

  • Sample standard deviation: $s = \sqrt{\frac{1}{N-1}\sum<em>{i=1}^N (x</em>i - \bar{x})^2}$

• Interpretation: a low standard deviation signals that most scores cluster around the mean; a high standard deviation signals wide dispersion.

Percentiles and standardized testing context

• Percentile definition: a percentile indicates the value below which a given percentage of observations fall.

• Example interpretation: a student scoring in the 90th percentile scores higher than 90% of peers who took the same test.

• Standardized testing context mentioned:

  • Percentiles and testing standards have historically been used (e.g., ACT, SAT) to rank or assess performance.

  • There was commentary about shifts in the usage and perceived role of these tests in college admissions.

Statistical significance, peer review, and replication

• Experimental results are evaluated by statistical significance: the probability that results are due to random chance rather than a real effect.

• If confounding variables are minimized or eliminated, the study is closer to a pure experiment.

• Key validity concepts:

  • Peer review: other experts read and critique the study or test to assess its quality and validity.

  • Replicability (replication): another class/school can reproduce the study with similar procedures and obtain close scores, demonstrating reliability across settings.

• The speaker referenced prior content about confounding (compounding) variables and emphasized the importance of valid, verifiable research practices.

Practical takeaways and context

• Data literacy takeaway: always consider how data are collected and presented (qualitative vs quantitative, scale type, and how bins/categories are defined).

• Be wary of marketing-driven data presentations and check the underlying definitions and time frames.

• Understand that correlation does not imply causation; use the correlation coefficient to gauge linear association, not to claim cause and effect.

• Recognize that outliers influence measures of central tendency and spread, and decisions about including or excluding them can change interpretations.

• In testing and assessment contexts, aim for a balanced distribution with reasonable central tendency and low to moderate variability to ensure fair measurements across populations.

Quick recap of key symbols and terms

• Central tendencies: mode, mean (\bar{x}), median

• Range: $ext{range} = \max(x<em>i) - \min(x</em>i)$

• Correlation: $r = \frac{\mathrm{cov}(X,Y)}{\sigma<em>X\,\sigma</em>Y}, \quad r \in [-1,1]$

• Standard deviation: $\sigma = \sqrt{\frac{1}{N}\sum<em>{i=1}^N (x</em>i - \mu)^2}$ (population) and $s = \sqrt{\frac{1}{N-1}\sum<em>{i=1}^N (x</em>i - \bar{x})^2}$ (sample)

• Percentiles: value below which a given percentage of observations fall

• Distributions: skewness (positive/negative), normal (bell curve), bimodal (two peaks)

• Significance concepts: statistical significance, peer review, replication