Biological Data Collection, Error, and Drawing Conclusions
Biological data collection means recording measurements or observations in an organized way, and drawing conclusions means using the processed data, while taking account of error and reliability, to decide whether a hypothesis is supported. In brief: qualitative data describe qualities, such as color, while quantitative data are numbers, such as mass in grams. Averages such as the mean and the range summarize results, and repeating measurements makes the results more reliable.
Living organisms vary, and measurements are never perfect, so careful handling of data is a core scientific skill. This guide explains types of data, how to build tables and graphs, how to calculate the mean and range, how errors, accuracy, reliability and reproducibility differ, and how to write a valid conclusion, with a worked example.
What are qualitative and quantitative data?
Qualitative data describe qualities that are not measured with numbers. Examples are the color of a leaf, the shape of a cell or the smell of a culture. They are useful but subjective unless a clear description or scale is used.
Quantitative data are numerical, and they are the main type for tests of hypotheses. They are either discrete, such as the number of leaves, which can only be whole numbers, or continuous, such as height or temperature, which can take any value within a range. Quantitative data should always include units and an appropriate number of decimal places, matching the precision of the instrument.
How do you present data in tables and graphs?
A results table should have a title, column headings with units in the heading and not in each cell, and the independent variable in the first column. Repeats are placed in separate columns, followed by the mean.
The type of graph depends on the data.
- Line graph: for two continuous variables, such as time and temperature. Points are joined by a line or a best-fit curve.
- Bar chart: when one variable is categorical, such as the species of plant.
- Histogram: for the frequency of continuous data in classes, for example the distribution of leaf lengths.
- Scatter graph: to look for a relationship between two measured variables.
The independent variable goes on the x-axis, and the dependent variable goes on the y-axis. Axes need labels with units, and the scale must be even. Plot points accurately and draw a best-fit line that passes near the points, but not through every point.
How do you calculate the mean and range?
The mean is the sum of the values divided by the number of values. The range is the difference between the largest and smallest value, and it is often given as the two end values. The median is the middle value when data are in order, and the mode is the most frequent value.
Anomalous results, which clearly do not fit the pattern, should be identified and either repeated or left out of the mean, with an explanation. Do not delete data without a reason.
What is experimental error?
All measurements differ from the true value by some amount. Two broad types of error are recognized.
- Random error changes the result unpredictably, sometimes higher and sometimes lower. Examples are small differences in reading a scale or natural variation between organisms. Repeating the measurement and taking a mean reduces its effect.
- Systematic error shifts all results in the same direction. Examples are a balance that is not zeroed, a thermometer that reads 2 °C too high or a stopwatch started late each time. Repeating does not remove it. It is corrected by calibrating the instrument or changing the method.
Mistakes by the experimenter, such as writing a number down incorrectly, are not treated as experimental error. They are blunders and should be avoided or corrected.
What do reliability, accuracy and reproducibility mean?
| Term | Meaning | How to improve it |
|---|---|---|
| Accuracy | Closeness to the true value | Calibrate instruments, correct systematic errors |
| Precision | Closeness of repeated measurements to each other | Use better instruments, careful technique |
| Reliability | Consistent results when repeated in the same way | More repeats, larger samples |
| Reproducibility | Same result obtained by different people or methods | Clear method, independent repeats |
Accuracy and precision are not the same. A scale that always reads 5 g too high gives precise but inaccurate results. A well-designed experiment aims for both.
Worked example: from raw data to a conclusion
A student tests the hypothesis that more sugar in the solution increases the rate of yeast fermentation. She measures the volume of gas in milliliters produced in 10 minutes, with three repeats for each sugar concentration.
| Sugar (%) | Trial 1 (mL) | Trial 2 (mL) | Trial 3 (mL) | Mean (mL) | Range (mL) |
|---|---|---|---|---|---|
| 0 | 1 | 2 | 1 | 1.3 | 1 to 2 |
| 2 | 8 | 9 | 10 | 9.0 | 8 to 10 |
| 4 | 17 | 15 | 16 | 16.0 | 15 to 17 |
| 6 | 22 | 23 | 24 | 23.0 | 22 to 24 |
- Means: (1 + 2 + 1) ÷ 3 = 1.3; (8 + 9 + 10) ÷ 3 = 9.0; (17 + 15 + 16) ÷ 3 = 16.0; (22 + 23 + 24) ÷ 3 = 23.0.
- Pattern: the mean volume increases as the sugar concentration increases, in each step by 7.7, 7.0 and 7.0 mL.
- Reliability: the ranges are narrow (2 mL or less) and do not overlap between concentrations, so the repeats agree well.
- Limit: only four concentrations were tested, so the data cannot show what happens at higher sugar levels, where the increase might stop.
Systematic errors to consider include reading the volume at the wrong point on the scale or keeping the water bath at a slightly different temperature for each tube. Random error is shown by the small differences between trials.
How do you write a valid conclusion?
A good conclusion has four parts: it states the relationship found, supports it with data, relates it to the hypothesis, and mentions limitations. For the yeast data: “The results support the hypothesis. As sugar concentration rose from 0 to 6 percent, the mean gas volume rose from 1.3 mL to 23.0 mL. The repeats were consistent, with ranges of 2 mL or less. The investigation tested only four concentrations, so the trend at higher concentrations is unknown.”
A conclusion should not claim more than the evidence shows. A correlation does not prove a cause unless other variables were controlled. Suggest improvements, such as more concentrations, more repeats or a thermostatically controlled bath.
How did statistics enter biology?
Francis Galton and Karl Pearson developed statistical tools for studying variation and inheritance in the late nineteenth century. Ronald Fisher, working at an agricultural research station in the 1920s, introduced methods for designing experiments and analyzing variance, which made it possible to separate real effects from random variation. Today statistical software is standard in biological research.
Which mistakes are common?
Mistake 1: Averaging anomalous results
An obvious outlier can distort the mean. Check it, repeat it if possible, and explain any decision to exclude it.
Mistake 2: Believing that repeats remove all error
Repeating reduces random error only. Systematic error requires a change in method or calibration.
Mistake 3: Using a line graph for categories
Joining points for categories such as species suggests a trend between them that does not exist. Use a bar chart.
Mistake 4: Writing a conclusion without data
“The hypothesis was correct” is not enough. Include values from the results and say how they support the claim.