Unit 22: Biostatistics — Long Questions

Mean The mean (average) is calculated as the sum of all values divided by the number of values (Mean = Σx/n for ungrouped data, or Σfᵢxᵢ/Σfᵢ for grouped data). It is most appropriate for symmetric data without extreme outliers — e.g., finding the average weight of birds of the same species from a sample (400, 420, 470, 390, 460 g gives a mean of 428 g), since it uses every value and gives a true 'typical' measurement.

Median The median is the middle value of data arranged in order (average of the two middle values if n is even). It is most appropriate when the data is skewed or contains outliers, such as population size, disease prevalence, or body measurements that are not normally distributed, because it is not distorted by extreme values — e.g., for the weights 300, 315, 330, 400, 450, 500 kg, the median is (330+400)/2 = 365 kg.

Mode The mode is the most frequently occurring value in a dataset. It is most appropriate for identifying the most common trait or characteristic in a biological sample — e.g., the most common plant height in a garden (50 cm appearing 4 times out of 10 measurements) — and is the only central-tendency measure that can be used for categorical data.

Standard deviation (σ) measures how dispersed data is in relation to the mean; it is calculated by finding the deviation of each data point from the mean, squaring these deviations, averaging them (variance), and taking the square root. A low SD means data clusters tightly around the mean; a high SD means greater spread. Advantage: it uses every data point, giving a precise and statistically robust picture of variability (e.g., useful for comparing variability in enzyme concentration or growth rate between samples). Limitation: it is more laborious to calculate and, like the mean, can still be influenced by extreme outliers.

Range is the simplest measure of spread, calculated as Range = Xmax − Xmin. Advantage: it is very quick and easy to calculate and interpret. Limitation: it only uses the two extreme values and ignores the distribution of all other data points, so it can be heavily distorted by a single unusually high or low value and gives no information about how data is spread between the extremes.

Percentiles indicate the value below which a given percentage of observations fall (e.g., the 25th, 50th and 75th percentiles, or quartiles). Advantage: they are particularly useful for understanding the distribution of skewed or non-normal biological data, and allow comparison of an individual's position relative to a reference population (e.g., growth charts). Limitation: percentiles describe position within the dataset rather than giving a single overall value of spread, and calculating them (with interpolation) is more involved than the range.

Overall, range is the quickest but least robust; SD is the most comprehensive and statistically useful for normally distributed data; percentiles are best for understanding distribution shape and skewed biological data.

Error bars are graphical elements added to bar charts (and other plots) to visually represent the variability or uncertainty in data, showing how far individual data points typically deviate from the central value (the mean) shown by the bar.

Types 1. Range error bars use the smallest (minimum) and largest (maximum) values in each group; the bar extends from minimum to maximum, showing the full spread of the data.
2. Standard deviation (SD) error bars extend above and below the mean by one (or more) standard deviations (Mean ± SD), showing the average distance of data points from the mean.

Significance Error bars let a reader judge, at a glance, how reliable or variable a reported mean is and whether differences between experimental groups are likely to be meaningful — large, overlapping error bars suggest the groups may not differ significantly, while small, non-overlapping error bars suggest a real difference.

Illustration (hypothetical dataset — bean plant height, cm, under different light conditions):
Group A (low light): 10, 11, 9 → Mean = 10 cm, Range = 9–11, SD ≈ 1
Group B (medium light): 14, 16, 15 → Mean = 15 cm, Range = 14–16, SD ≈ 1
Group C (high light): 18, 20, 19 → Mean = 19 cm, Range = 18–20, SD ≈ 1
On the bar chart, bars are drawn at heights 10, 15, and 19 cm for groups A, B and C. SD error bars would extend ±1 cm around each bar top (e.g., Group A from 9 to 11), and range error bars would extend from each group's minimum to its maximum value, with horizontal caps added at each end for clarity.

Bar chart A graphical representation that uses rectangular bars to display categorical data, where the length/height of each bar is proportional to the value it represents. Categories are placed on the x-axis and values/frequencies on the y-axis; bars are of equal width and evenly spaced. Example construction: in an insect diversity study, the number of individuals of five species (Ladybug 45, Ant 120, Butterfly 30, Bee 75, Dragonfly 20) is plotted with insect species on the x-axis and number of individuals on the y-axis (scaled 0–140), producing a chart titled 'Insect species abundance in local forest'.

Pie chart A circular statistical graph divided into sectors, where the whole circle represents 100% (360°) and each sector's angle is proportional to that category's share of the total (Sector angle = category value/total value × 360°). Example construction: for 100 students' blood types (A=30, B=20, AB=10, O=40), sector angles are A=108°, B=72°, AB=36°, O=144°, drawn as 'Distribution of blood types among students'.

When to favour each Bar charts are favoured when comparing absolute counts or magnitudes across several distinct categories or groups, especially when there are many categories or when exact comparison of sizes matters (e.g., comparing counts of different insect species, or comparing experimental group means). Pie charts are favoured when the goal is to show how a few categories contribute proportions of a single whole (e.g., showing what percentage of a student population has each blood type, or the proportion of different cell types in a sample) — they become hard to read if there are too many categories or if proportions are very close in size, a situation where a bar chart communicates the comparison more clearly.

Bar chart Displays categorical data using rectangular bars whose height is proportional to the value represented, with categories on the x-axis and values on the y-axis. Best used to compare quantities across distinct categories — e.g., comparing the number of individuals of five insect species (Ladybug, Ant, Butterfly, Bee, Dragonfly) counted in a forest survey.

Pie chart A circular chart where the full circle (360°, 100%) is divided into sectors proportional to each category's share of the total. Best used to show how parts make up a whole — e.g., showing the distribution of blood types (A, B, AB, O) among 100 surveyed students.

Error bars Graphical additions to a bar chart (range bars or SD bars) that show the variability or uncertainty around a group's mean. Best used whenever an experiment compares means of several groups and the researcher needs to show how reliable or variable those means are — e.g., showing the mean plant height (with SD or range error bars) of bean plants grown under low, medium, and high light conditions, so readers can judge whether light intensity caused a real difference in growth.

1. Question: Does the new drug reduce recovery time from flu compared to standard care?
2. Hypothesis: Patients who receive the drug will recover from flu faster than those who do not.
3. Variables: Independent variable — whether a patient receives the drug or not; Dependent variable — recovery time (number of days until symptoms resolve); Control variables — patient age, sex, baseline severity of illness, any other medication, and the timing/method of recovery assessment, all kept the same across groups.
4. Groups: Control group — flu patients given a placebo or standard supportive care (no drug); Experimental group — flu patients given the new drug.
5. Procedure: Select flu patients matched for age and baseline severity; randomly assign them to the control and experimental groups to avoid bias; if possible, use a double-blind design (neither patient nor assessor knows the group) to prevent bias in reporting/measuring recovery; administer the assigned treatment under identical conditions; record each patient's recovery time using the same criteria for all patients; use an adequate sample size so results are not due to chance.
6. Analysis: Compare the mean recovery time (and its standard deviation) between the two groups using appropriate statistical tests to see if the difference is significant.
7. Conclusion: If the drug group shows a statistically significant reduction in mean recovery time compared to the control group, the hypothesis is supported; if not, the hypothesis is rejected.

Validity is ensured through randomization, use of a control group, blinding, and controlling confounding factors (age, severity, concurrent treatment) so that only the drug's effect is being tested. Reliability is ensured through an adequately large sample size, a standardized procedure applied identically to all subjects, and repeatability of the protocol.

The range (Xmax − Xmin) is the simplest and quickest measure of spread, but it uses only the two most extreme values in a dataset and completely ignores how the remaining data points are distributed. This makes it highly sensitive to a single unusually high or low value (an outlier), which can give a misleading impression of the overall variability of a sample.

The standard deviation, by contrast, is calculated using the deviation of every single data point from the mean, squaring and averaging these deviations before taking the square root. Because it incorporates all observations, it gives a far more representative and statistically robust picture of how spread out the data is around the mean.

In medical research, where decisions about drug safety, treatment effectiveness, or disease risk affect patient care, I would trust the standard deviation more, because it reflects the behaviour of the whole sample rather than just its two extremes, making it less likely to be distorted by one atypical patient's result and more useful for comparing variability between treatment groups.

During an epidemic, mortality data is often skewed — a few regions, age groups, or time periods may show very high mortality while most others are much lower. If policymakers rely only on the mean mortality rate, a small number of extreme values can pull the mean higher or lower than what is 'typical', giving either a falsely alarming or falsely reassuring picture of the epidemic's severity. The median, being the middle value, would better represent the 'typical' mortality experience and is less affected by extreme outlier regions, but using the median alone could cause policymakers to underestimate how severe the worst-affected areas really are. Percentiles can show, for example, that the worst 10% of regions (the top 90th percentile) have a much higher mortality rate than the rest, highlighting at-risk populations — but if policymakers misread a percentile value as applying to everyone, they may design interventions that don't target the areas that need them most.

Biostatistics prevents such misinterpretation by presenting several complementary measures together — mean, median, percentiles, and a measure of spread such as range or standard deviation — rather than a single number, and by clearly communicating the context, limitations, and assumptions behind the analysis, as emphasized in the 'Interpretation of Results' step of biostatistical analysis. This ensures policymakers see both the overall/typical trend and the extent of variation or at-risk subgroups, enabling fairer and more targeted public health decisions.