Comprehensive Study Notes on Statistical Significance, Data Classification, and Levels of Measurement

Statistical Significance

Definition: Data has statistical significance when an outcome is extremely unlikely to occur purely by chance. The standard standard baseline threshold for statistical significance is a probability of occurrence of less than $5\%$ ( $P < 0.05$ ).
Normal Range and Outliers:
- A normal distribution or standard benchmark establishes a lower limit and an upper limit for normal variation.
- Data falling outside these limits are termed outliers or extremes. These terms are objective statistical descriptions and carry no qualitative or value judgments.
- Both the extremely low end (below the lower limit) and the extremely high end (above the upper limit) represent outcomes with less than a $5\%$ chance of happening.
- Example: In standard IQ scoring, the normal range is from $70$ to $115$ .
- An IQ score above $115$ is classified as extremely high.
- An IQ score below $80$ is classified as significantly low and holds statistical significance.
- If $100$ people are randomly selected, fewer than $5$ individuals ( $5\%$ ) are expected to have IQ scores falling into either extreme tail.
Case Study: Pea Hybrid Genetics Experiment:
- Theoretical Expectation: According to genetic theory, $25\%$ of pea plants should yield yellow pods.
- Sample Size: $504$ total peas.
- Expected Value Calculation: $\text{Expected Yellow Peas} = 504 \times 0.25 = 126$
- Observed Result: The experiment yields $129$ yellow pods out of $504$ ( $\approx 25.6\%$ ).
- Data Under Study: The data of interest in statistical significance testing is the difference between observed reality and theoretical expectation. $\text{Observed Difference} = |129 - 126| = 3$
- Significance Threshold Calculation: $\text{Threshold} = 504 \times 0.05 = 25.2$
- Because whole discrete peas are being counted, $25.2$ is rounded up to $26$ .
- Significance Evaluation:
- To be statistically significant, the absolute difference between the theoretical expectation ( $126$ ) and the observed reality must be equal to or greater than $26$ (representing the $5\%$ significance threshold).
- Since the observed difference is only $3$ ( $3 < 26$ ), the result is not statistically significant.
- An observed difference of $129$ or any difference of $26$ or greater would be considered statistically significant, as such a large deviation from the theoretical model is extremely unlikely to occur by chance.

Classification and Characterization of Data

Data can be categorized simultaneously through two distinct, non-contradictory structural frameworks:

Qualitative (Categorical) vs. Quantitative (Numerical) Data.
The Four Levels of Measurement.

Qualitative / Categorical Data:
- Consists of descriptive categories, names, labels, or non-numerical attributes.
- Cannot be measured using a scale or ruler; it is analyzed by counting occurrences or frequencies.
- Categorical Examples:
- Gender (even if coded numerically as $1 = \text{female}$ and $2 = \text{male}$ , the numbers serve solely as labels without mathematical magnitude).
- Eye color and hair color.
- Personal opinions, product preferences, political candidate choices, and social media activity.
- Zip Codes (despite consisting of digits, zip codes represent geographic identification labels rather than measured quantities).
- Social Security Numbers (numerical identifiers assigned to individual identity rather than measured values).
Quantitative / Numerical Data:
- Consists of numerical values representing measured quantities, counts, or amounts that carry mathematical magnitude and units of measurement.
- Categorized into two sub-types: Discrete and Continuous.
- Discrete Data:
- Consists of finite, countable values.
- Characterized by integer values without meaningful fractions between adjacent units.
- Discrete Examples:
  - Number of desks in a classroom.
  - Number of cars in a lot.
  - The number of times a $2$ is observed when rolling a die $15$ times.
  - Number of days.
  - Currency/Money (although money uses two decimal places, it remains discrete because it is countable in fixed, non-divisible minimum increments of $0.01\,\text{dollars}$ or $1\,\text{cent}$ ).
- Continuous Data:
- Consists of infinitely countable values along a continuum.
- Between any two given numbers, there is always an infinite number of intermediate values (e.g., between $2.5$ and $2.6$ , there exists $2.55$ ).
- Dependent on the precision of the measuring instrument; characterized by potential decimal expansions.
- Continuous Examples:
  - Volume of water in a cup (measured in gallons, liters, or $\text{cm}^3$ ).
  - Body weight of an individual or infant (measured in ounces, pounds, kilograms, or grams).
  - Temperature.
  - Length of rainbow chalk (measured in meters, centimeters, or millimeters).
Age Classification Exception:
- Age is strictly Quantitative / Numerical (Continuous), not categorical. Age is a measured duration of time elapsed since birth (where $0\,\text{years}$ represents an absolute baseline origin), calculated as multiples of $365\,\text{days}$ .

Four Levels of Measurement

Data is further organized into four hierarchical levels of measurement (detailed across pages 17 to 19 of standard statistical frameworks):

Nominal Level: Categories, labels, or names only. No natural ordering or mathematical operations apply (e.g., hair color, zip code).
Ordinal Level: Data can be arranged in a specific order or ranking, but differences between values are meaningless or cannot be calculated (e.g., satisfaction rankings).
Interval Level: Ordered data where differences between values are meaningful, but there is no absolute natural zero baseline (e.g., temperature in Fahrenheit or Celsius).
Ratio Level: Ordered data with meaningful differences and a true, absolute zero baseline where zero indicates the total absence of the quantity (e.g., height, weight, volume, age).

Methods for Analyzing Categorical Data

Categorical attributes (e.g., colors like red, blue, or gray) cannot be mathematically added or averaged ( $\text{blue} + \text{red}$ divided by $2$ is undefined).
Analytical Method:
1. Determine the absolute count or frequency of occurrences ( $n$ ) for each discrete category within a sample.
2. Describe and analyze the categorical distribution using proportions or percentages of the total sample size.
3. Discrete integers are used to quantify and evaluate underlying categorical data distributions.

Questions & Discussion

Question: Is age considered a categorical data type?
- Answer: No. Age is measured starting from an absolute zero point ( $0\,\text{years}$ ) in increments of days and years. Therefore, it is a quantitative measurement, not categorical.
Question: Does an IQ score or data point outside normal limits imply a qualitative judgment?
- Answer: No. Terms like "outlier" or "extreme" are strictly structural descriptions of statistical probability indicating that an observation falls in the lower or upper tails ( $< 5\%$ chance of occurrence).
Question: In the pea hybrid experiment ( $504$ total peas, theoretical expectation $25\%$ yellow pods), how is statistical significance evaluated step-by-step?
- Answer:
1. Theoretical expected count: $504 \times 0.25 = 126$ .
2. Actual observed count: $129$ yellow pods.
3. Measured data point (difference): $|129 - 126| = 3$
4. Threshold for significance ( $5\%$ of sample size): $504 \times 0.05 = 25.2$ , which rounds up to $26$ discrete units.
5. Because the difference $3$ is less than the threshold $26$ , the deviation is within normal expected variance and is not statistically significant.
Question: Would a result be statistically significant if the observed number of yellow pods differed from expectation by $26$ or more?
- Answer: Yes. If the difference between the observed number and the expected baseline ( $126$ ) is $26$ or greater, the outcome crosses the $5\%$ probability threshold and is deemed statistically significant.
Question: Why are decimal values usually continuous, while money is treated as discrete?
- Answer: Continuous variables allow infinite fractional precision depending on measuring accuracy (e.g., volume or weight). Money is discrete because financial transactions are limited to a fixed minimum indivisible unit ( $0.01\,\text{dollars}$ or $1\,\text{cent}$ ), making currency strictly countable.