Designing a STUDY

• Designing a Study:

  • Ability to generate hypotheses that can be falsified through observation and experiment
  • Involves questions, hypotheses, predictions, nature of data to collect, how much data to collect, how to collect data, experimentation (control), and statistical analyses
• Aim:

  • Purpose of the study

  Objective:

  • Basic approach to achieve the aim(s), clear, concise, and attainable

  Hypothesis:

  • Clear statement articulating a plausible explanation for observations, allows gathering data to refute or support it

  Prediction(s):

  • Expected observations in the experiment or further observation
• Three Approaches to Scientific Investigation:

  • Descriptive Studies: Based on observations, describes scientific value without relating variables or determining cause and effect
  • Correlational Studies: Examine relationships between variables using natural data and variation, no manipulation by the investigator
  • Experimental Studies: Relate variables by manipulating the independent variable within the natural range of variability for the organism/system
• Experimental Studies:

  • Treatments include the whole natural range of the independent variable
  • Control: A replica of the experimental treatment without manipulating the independent variable, used to attribute differences to the treatment
  • Factors to consider: Randomisation, replication, independence
• Types of Control:

  • "Before and after, fully controlled" method
  • "Before and after, without control group" method
  • "After only" method
  • Some studies may have no control
• Replication:

  • Set or group of samples manipulated and measured in the same way
  • Should be at least three for a reliable estimate of the mean/relationship
  • Pseudo-replication occurs when replicates/samples are not independent
• Randomisation:

  • Assignment and handling of samples and treatment groups to reduce bias
  • Each member of a population must have an equal and independent chance of being sampled
  • Methods: Computer-generated table of random numbers, flipping a coin
• Sampling:

  • Important to have a large sample size evenly spread across the whole range for a confident estimate of reality
  • Consider type of data, target species, size, distribution, behavior before deciding on a method
• The ability to generate hypothesis that can potentially be falsified through observation and experiment
• Designing a study


  1. Question

  2. Hypothesis

  3. Prediction

  4. Nature of data to collect

  5. How much data to collect

  6. How to collect the data

  7. Experimentation (control?)

  8. Statistical analyses
• Aim - Purpose of the study
• Objective


  The basic approach you plan to take to achieve your aim(s). Must be clear, concise and attainable.
• Hypothesis


  Clear statement articulating a plausible, possible explanation for observation(s). Allows gathering of data that can be used to refute or support it.
• Prediction(s)


  What you expect to observe in the experiment/further observation.
• There are THREE approaches to scientific investigation: Descriptive studies, Correlational studies, and Experimental studies
• Descriptive studies


  • Based on observations, describes something of scientific value, no attempt to relate variables or determine cause & effect
• Correlational studies


  • Examine relationships between variables, uses natural data and variation, no manipulation by investigator
• Experimental studies


  • Relate variables by manipulating the independent variable, experimentation (Manipulate & Control), still within the natural range of variability for the organism/system
• Treatments in experimental studies should include the whole natural range of independent variable, biological relationships are rarely linear, have enough treatments for relationship to be clearer, choose treatments carefully
• Control


  A replica of experimental treatment but without manipulating the independent variable, all other conditions must be the same & measured the same way, differences between experimental treatments and the control can be directly attributed to the treatment (independent variable), controls for confounding variables (hidden treatments)
• Types of control


  • "Before and after, fully controlled" method
  • "Before and after, without control group" method
  • "After only" method
• Some studies have no control, controls are associated with experimental studies but can also be included in correlational studies
• Factors to be considered in study design

  • Randomisation
  • Replication
  • Independence
• Replicates


  A set or group of samples that are manipulated & measured in same way, independent of each other, should be at least three (3) for reliable estimate of the mean/relationship, if replicates / samples are not independent, this is called pseudo-replication and conclusions will not be valid
• Randomisation


  Random data collection: Assignment and handling of samples and treatment groups, reduces bias, each member of a population must have equal & independent chance of being sampled, statistical procedures assume that samples are obtained randomly
• Adequate sampling is important, collecting and measuring a large sample, sample = all individuals in a specific treatment, must be large enough to give a confident estimate of population, relationship studies: Collect a large # samples to cover range of variability for individual variables
• Need large sample size evenly spread across the whole range for confident estimate of reality, know the range of your independent variable, replicate, over sampling – no increased return (in terms of accuracy & confidence in results)
• Sampling methods


  Consider type of data, target species, size, distribution and behaviour, know your study animal/plant/system before deciding on method
• Sampling in the field

  • Censusing Plants and Animals
  • Line transects
  • Distance Sampling
  • Mark-recapture sampling/Lincoln-Pearson Method
  • Point transect
  • Frame quadrats
  • Line Intercept
  • Nearest-individual (nearest-neighbour) method
• Measurements are important in quantitative studies, quantitative data meaningless without units, density measurement = # individual per unit area
• Accuracy & Precision


  Confidence or uncertainty of a measurement
• Sources of bias/error in measurements


  • Mobile animals – easy to count some twice, or miss some
  • Cryptic animals / plants
  • Instrument error / inaccuracy
  • Human error – especially when more than one person is collecting the data
• Ethical considerations: No abuse, no distress of animals, do you have to kill? Again!!, study affects environment?, study affects species dynamics?, medical trials are especially sensitive
• Scientific Method has limitations: Failures through prejudice, poor experimental design, misinterpretation of results, inadequate data collection, can we truly measure intelligence?