l stoodialinza
← All topics

Step 1 · Basics · Topic 1 of 17 · 5 min

Data: observations and variables

Before calculating anything, decide what one observation in the table is and what type each quantity has.

In short#

  • The observational unit defines n. Repeated measurements of one object are not independent observations (pseudoreplication).
  • A variable may be quantitative (continuous or discrete), ordinal or nominal. Its type affects which summaries and analysis methods are appropriate.
  • A missing value is not zero. Exclusion preserves the recorded value and makes the analysis decision reviewable.
  • Data structure determines the table type: numerical groups, X and Y pairs, factors with repeated measurements, time to event or category counts.

In plain words#

Imagine measuring plant height. Each plant is one observation. Height in centimetres is a variable whose value can differ between plants.

Variables have different types. Height is measured numerically. Flower color is recorded as a category. Ratings such as poor, fair and good can be ordered, but the gaps between levels are not necessarily equal. Variable type affects which summaries and analyses are appropriate.

Measuring one plant three times does not mean that you studied three plants. Repeated measurements help assess random measurement error, but they do not increase the number of independent plants.

In Linza#

  1. Choose a table type by data structure: Column for numerical groups, XY for pairs, Grouped for two factors and repeats, Survival for time to event, Contingency for counts.
  2. An empty cell is missing. To remove a value from analysis without deleting it, use exclusion (×).
  3. For repeated measurements, set an experimental-unit ID in the Factor data table.
Try it in Linza Opens in a new tab with synthetic data. Work in the original tab remains unchanged.

Common mistakes#

  • Technical replicates treated as independent observations: the sample size is inflated and p-values may be too small.
  • Missing values replaced with zeros: the mean may be biased.
  • Percentages entered as frequencies in a contingency table.

Check yourself#

You weighed 4 independent mice 3 times each. How many independent observations are there in this example?

Optional. Your answer is not saved.

Sources#

  • Hurlbert S.H. (1984) Pseudoreplication and the design of ecological field experiments. Ecological Monographs, 54(2). Search for publication Opens in a new tab
  • Lazic S.E. (2010) The problem of pseudoreplication in neuroscientific studies: is it affecting your analysis? BMC Neuroscience, 11. Open publication Opens in a new tab

References are for further reading. A method’s suitability depends on the study design and its assumptions.