Distribusi Normal

01Core Concept and Definition

The normal distribution is a continuous probability distribution that forms a symmetric, bell-shaped curve when plotted. It is the single most important distribution in biostatistics because the majority of biological measurements in large populations (blood pressure, height, serum lab values, birth weight) tend to follow this pattern.
The curve is defined entirely by two parameters: the mean ($\mu \backslash mu$), which sets the center of the distribution, and the standard deviation ($\sigma \backslash sigma$), which controls how spread out the data points are. A smaller $\sigma \backslash sigma$ produces a tall, narrow curve clustered tightly around the mean; a larger $\sigma \backslash sigma$ produces a flatter, wider curve.
Three properties make the normal distribution recognizable on exam questions. First, the mean, median, and mode are all equal and sit at the exact center of the curve. Second, the distribution is perfectly symmetric, meaning the left and right halves are mirror images. Third, the tails of the curve extend infinitely in both directions but never touch the x-axis, approaching zero asymptotically.
When a question describes a variable as "normally distributed" or shows a bell-shaped histogram, you should immediately think of these properties because the question will almost certainly test your ability to apply the empirical rule or calculate a Z-score.

02The Empirical Rule (68-95-99.7 Rule)

This rule is the most heavily tested property of the normal distribution. If a vignette tells you the mean systolic blood pressure of a population is 120 mmHg with a standard deviation of 10 mmHg, you must be able to state that approximately 68% of the population will have systolic blood pressures between 110 and 130 mmHg, 95% between 100 and 140 mmHg, and 99.7% between 90 and 150 mmHg.
The 95% range ($\mu \pm 2\sigma \backslash mu\; \backslash pm\; 2\backslash sigma$) is particularly important because it directly underpins the concept of the 95% confidence interval. In practice, the precise multiplier for 95% is 1.96 standard deviations, but for rapid calculation on the exam, using 2 is sufficient and expected. Understanding that 5% of data lies outside $\mu \pm 2\sigma \backslash mu\; \backslash pm\; 2\backslash sigma$ also means that 2.5% lies in each tail, which connects directly to the concept of a two-tailed hypothesis test with $\alpha =0.05\backslash alpha\; =\; 0.05$.
A common exam setup gives you the mean, the standard deviation, and a sample size, then asks what percentage of individuals fall above or below a certain value. To answer, determine how many standard deviations that value is from the mean, then apply the empirical rule.

03Z-Scores and the Standard Normal Distribution

The Z-score converts any normally distributed variable into a standard normal distribution, which has a mean of 0 and a standard deviation of 1. This standardization allows you to compare values from completely different scales. For example, you can compare a patient's hemoglobin level to their white blood cell count by converting both to Z-scores.
A Z-score of +2.0 means the value is 2 standard deviations above the mean. A Z-score of -1.5 means the value is 1.5 standard deviations below the mean. In the context of hypothesis testing, a Z-score beyond $\pm 1.96\backslash pm\; 1.96$ corresponds to a p-value less than 0.05 (two-tailed), which leads to rejecting the null hypothesis.
The distinction between $\sigma \backslash sigma$ and $SEMSEM$ is a classic exam target. The standard deviation describes the spread of individual data points in a population. The standard error describes the spread of sample means if you were to repeatedly draw samples of size $nn$ from the population. As $nn$ increases, SEM decreases (because of the $\sqrt{n}\backslash sqrt\{n\}$ in the denominator), meaning your estimate of the population mean becomes more precise. This is why larger studies produce narrower confidence intervals.
The confidence interval for the mean is calculated as:
$$CI=\bar{X}\pm Z\times SEMCI\; =\; \backslash bar\{X\}\; \backslash pm\; Z\; \backslash times\; SEM$$
For a 95% confidence interval, $Z=1.96Z\; =\; 1.96$. This formula appears frequently on exams and connects the Z-score, sample mean, and standard error into a single testable concept.

04Central Limit Theorem

The Central Limit Theorem (CLT) is the reason the normal distribution dominates biostatistics. Even if individual patient values follow a skewed or irregular distribution (for example, hospital length-of-stay data, which is right-skewed), the means of repeated random samples drawn from that population will form a normal distribution as long as the sample size is large enough.
This theorem justifies the use of Z-tests and t-tests in research. Without it, parametric statistical methods would only be valid for data that are already normally distributed, which is rarely the case in clinical medicine. The CLT is also why increasing sample size improves statistical power: the sampling distribution of the mean becomes narrower (smaller SEM) and more normal, making it easier to detect a true difference.
Exam questions on the CLT typically present a non-normal population distribution and ask what happens to the distribution of sample means as sample size grows. The answer is always that it becomes approximately normal, centered at the population mean, with a spread described by the SEM.

05Common Distortions: Skewness and Kurtosis

Recognizing skewness is frequently tested as a contrast to the normal distribution. The key rule is: the mean is pulled toward the tail. In a right-skewed distribution, a few extremely high values drag the mean to the right, making it larger than the median. In a left-skewed distribution, a few extremely low values pull the mean to the left.
When data are skewed, the median becomes a better measure of central tendency than the mean because the median is resistant to outliers. This is why clinical studies reporting skewed data (such as length of stay or cost) typically report the median rather than the mean.
Kurtosis describes the "peakedness" of a distribution. A leptokurtic distribution has heavier tails and a sharper peak than the normal curve; a platykurtic distribution has lighter tails and a flatter peak. Kurtosis is far less commonly tested than skewness, but understanding that it describes tail behavior (not just peak height) can help on occasional questions about the shape of distributions.
For exam purposes, the most important takeaway is that when a question describes data as "normally distributed," it is confirming that the mean equals the median equals the mode, and parametric tests are appropriate. When data are described as "skewed," use the median as your central measure and consider nonparametric tests.

06Applications in Medicine and Research

The concept of reference ranges is a direct clinical application. When a lab report says the normal range for serum sodium is 135 to 145 mEq/L, that range was derived by measuring sodium in a large group of healthy individuals, calculating the mean and standard deviation, and defining "normal" as $\mu \pm 2\sigma \backslash mu\; \backslash pm\; 2\backslash sigma$. This means that by definition, 5% of healthy individuals will have values outside the normal range (2.5% above, 2.5% below). This is a frequently tested pearl: a value outside the reference range does not automatically indicate disease, because the range itself excludes 5% of healthy people.
In screening and diagnostic testing, the normal distribution explains the tradeoff between sensitivity and specificity. If you lower the cutoff for a positive test (move the threshold to the left on the distribution), you capture more true positives (higher sensitivity) but also more false positives (lower specificity). This concept is tested alongside ROC curves.

07Traps and High-Yield Pearls

The most common mistake students make with normal distribution questions is confusing standard deviation with standard error of the mean. A question that asks about the spread of individual patient values is asking about $\sigma \backslash sigma$. A question that asks about the precision of the sample mean or about confidence intervals is asking about $SEM=\sigma \mathrm{/}\sqrt{n}SEM\; =\; \backslash sigma\; /\; \backslash sqrt\{n\}$. Mixing these up leads to incorrect confidence interval calculations and wrong answers about sample size effects.
A second trap involves the empirical rule percentages for one tail versus two tails. Students correctly recall that 95% of data falls within $\pm 2\sigma \backslash pm\; 2\backslash sigma$, but then forget that the remaining 5% is split between both tails. When a question asks for the probability of a value being above $\mu +2\sigma \backslash mu\; +\; 2\backslash sigma$, the answer is 2.5%, not 5%. This distinction is directly linked to understanding one-tailed versus two-tailed hypothesis tests.
A third common error is assuming that the Central Limit Theorem makes individual data points normal. It does not. The CLT applies only to the distribution of sample means, not to the raw data. Individual observations remain in whatever distribution they originally followed.
Finally, watch for questions that present a skewed histogram and ask for the best measure of central tendency. If the data are skewed, the answer is the median, not the mean. If the data are normally distributed, the mean, median, and mode are interchangeable, and the mean is preferred because it uses all data points.
The core competency being tested across all normal distribution questions is whether you can connect the shape of a distribution to the correct statistical approach: choosing the right summary statistic, applying the right formula, and interpreting the result within clinical context.