Descriptive and Inferential Statistics in Psychological Research

Descriptive and Inferential Statistics in Psychological Research

Psychology relies on quantitative measurement to systematically collect, summarize, analyze, and interpret empirical research data. Statistics are divided into two primary branches: descriptive statistics, which organize and summarize data, and inferential statistics, which allow researchers to draw conclusions and test hypotheses about population characteristics based on sample data.

---

Scales of Measurement (NOIR)

Variables in psychological research are measured using four fundamental scales of measurement, remembered by the acronym NOIR:

1. Nominal Scale: Categorical data with no inherent numerical order or quantitative ranking (e.g., eye color, psychiatric diagnoses, experimental condition labels).
2. Ordinal Scale: Data arranged in a distinct rank order or sequence, but where intervals between ranks are unequal or unknown (e.g., birth order, class rankings, finishing order in a marathon).
3. Interval Scale: Numerical data with equal intervals between adjacent values, but lacking a true or absolute zero point (e.g., Temperature in Fahrenheit or Celsius, standard IQ scores). A score of 0 does not mean the absence of the trait.
4. Ratio Scale: Numerical data featuring equal intervals and a true, absolute zero point representing the complete absence of the measured variable (e.g., reaction time in milliseconds, weight, height, score on a 100-item test). Ratio data allows for meaningful multiplicative statements (e.g., 20 seconds is twice as long as 10 seconds).

---

Descriptive Statistics: Central Tendency and Skewed Distributions

Measures of Central Tendency summarize an entire dataset with a single representative score:

• Mean: The arithmetic average calculated by summing all scores and dividing by the total number of scores ($N$). The mean is highly sensitive to extreme outlier scores.
• Median: The exact middle score when data are arranged in ascending order. If $N$ is even, it is the average of the two middle scores. The median is resistant to extreme outliers, making it the ideal measure for skewed distributions.
• Mode: The most frequently occurring score in a dataset. A dataset can be unimodal, bimodal (two modes), or multimodal.

Skewed Distributions

In a perfectly symmetrical Normal Distribution, the Mean, Median, and Mode are identical and sit at the exact center of the curve.

When extreme outlier scores pull the distribution to one side, the dataset becomes skewed:

• Positively Skewed Distribution (Right-Skewed): Contains a cluster of low scores with a long tail stretching toward the high positive end (e.g., household income data). The extreme high scores pull the Mean to the right. Relationship: Mean > Median > Mode.
• Negatively Skewed Distribution (Left-Skewed): Contains a cluster of high scores with a long tail stretching toward the low negative end (e.g., an extremely easy exam where most students score near 100%). The extreme low scores pull the Mean to the left. Relationship: Mode > Median > Mean.

---

Measures of Variability and Standard Scores

Measures of Variability (Dispersion) describe how spread out or clustered scores are around the central tendency:

• Range: The difference between the highest and lowest scores in a distribution. It is crude and heavily influenced by single extreme values.
• Variance ($s^2$): The average of the squared deviations of each score from the dataset mean.
• Standard Deviation ($SD$ or $s$): The square root of the variance. $SD$ reflects the average distance of scores from the mean in the original units of measurement.
• Z-Score (Standard Score): Expresses how many standard deviations a raw score ($X$) lies above or below the population mean ($\mu$):

$z = \frac{X - \mu}{\sigma}$

A $z$-score of $+1.5$ indicates the raw score is $1.5$ standard deviations above the mean.

---

Correlation and Bivariate Analysis

Correlation measures the strength and direction of a linear relationship between two quantitative variables, expressed by Pearson's Correlation Coefficient ($r$), which ranges strictly from $-1.00$ to $+1.00$:

• Direction: Indicated by the sign ($+$ or $-$):
  • Positive Correlation ($r > 0$): As one variable increases, the second variable also increases (e.g., study hours and exam scores).
  • Negative Correlation ($r < 0$): As one variable increases, the second variable decreases (e.g., missed classes and exam scores).
• Strength: Indicated by the absolute value of $r$ ($|r|$), independent of its sign. An $r = -0.85$ reflects a stronger relationship than an $r = +0.50$.
• Zero Correlation ($r = 0.00$): No linear relationship exists between the two variables.

CRITICAL RULE: Correlation Does NOT Imply Causation

Finding a strong correlation between Variable A and Variable B does not prove that A causes B. Three possibilities exist: A causes B, B causes A, or an unmeasured Third Variable (Confound) causes both A and B (e.g., ice cream sales and drowning rates correlate strongly because both are driven by hot summer temperatures).

---

Inferential Statistics and Hypothesis Testing

Inferential statistics allow researchers to determine whether patterns observed in a sample represent genuine effects in the broader population, or if they are simply random sampling error.

Hypothesis Testing Framework

• Null Hypothesis ($H_0$): The baseline assumption that there is no real effect, difference, or relationship in the population (any observed sample difference is due to chance).
• Alternative Hypothesis ($H_1$): The research hypothesis stating that a real effect, difference, or relationship exists.

Decision Matrix and Errors in Hypothesis Testing

Researchers evaluate sample data to either Reject $H_0$ or Fail to Reject $H_0$, creating two potential errors:

• Type I Error (Alpha Error): Rejecting a true null hypothesis (finding a false positive—claiming a drug works when it does not).
• Type II Error (Beta Error): Failing to reject a false null hypothesis (finding a false negative—failing to detect a drug that actually works).
• Statistical Significance ($p$-value): The probability that the observed sample results occurred by random chance under the assumption that $H_0$ is true. In psychology, the standard threshold is $p < .05$ (less than a 5% probability of chance occurrence).
• Effect Size (Cohen's d): Measures the quantitative magnitude of an experimental effect, independent of sample size.

---

Statistics Reference Table