Lecture 2: Ecology and The
Scientific Method
Understanding the world around us
•
•
Traditional environmental knowledge:
decisions based on personal experience,
observation, and received wisdom.
Experimental: posing a question
(hypothesis), collecting appropriate data
using an appropriate design and analysis
and converting a statistical result
back into a biological conclusion.
How Ecologists Study the
Natural World
• Ecologists, like other scientists, employ a
scientific method:
– observation and description of
natural phenomena
– development of hypotheses or explanations
– testing the predictions of these hypotheses
• We test hypotheses because many
explanations are plausible. Which is best?
1
What is an hypothesis?
• A hypothesis is an idea about how the
world works:
– e.g., “Frogs sing on warm nights after periods
of rain.”
• We often wish to understand two
components of such a phenomenon:
– how? (encompasses physiological
processes)
– why? (encompasses costs and benefits of the
behavior to the individual)
Experiments test predictions.
• Hypotheses generate predictions:
– if observations confirm the prediction,
the hypothesis is strengthened (not proven)
– if observations fail to confirm the prediction, the
hypothesis is weakened (or rejected)
• Best tests of hypotheses are experiments:
– independently manipulate one/few
variables
– establish appropriate controls
Ecology Employs the Scientific
Method
induction
induction
observation
hypothesis
or model
experiment
prediction
deduction
deduction
2
Some Potential Pitfalls
• A correlation between variables does not
establish causation.
• Many hypotheses cannot be tested by
experimental methods
because:
– the scale is too large:
• patterns may have evolved over long periods
• the spatial extent is too large for manipulation
– causal factors cannot be independently tested
Some Approaches to Difficult
Problems
• Microcosms are
sometimes useful:
– microcosms replicate
essential features of
the system in a
laboratory or field
setting
Some Approaches to Difficult
Problems
• Mathematical models are powerful tools:
– researcher portrays system as set of
equations
– model is an hypothesis and yields predictions
that can be tested; examples include:
• models of disease spread
• models of global carbon
3
Types of ecological evidence
• Observations of the natural world
• Laboratory experiments
• Mensurative experiments
• Manipulative field experiments
• Mathematical models
Classification of methods used in ecological
studies
1. Mensurative experiments: involve making
measurements at one or more points in space or
time. Space or time is the only experimental variable
or treatment.
Study of Uncontrolled events
Distinct perturbation
occurs
No distinct
perturbation occurs
Observational Studies
2. Manipulative experiments: involve two or more
treatments, and has as its goal, the making of more
comparisons.
– Many manipulative experiments can render ambiguous results unless
care is taken in experimental design.
Events controlled by observer
Replicated
Experiments
Unreplicated
Experiments
Sampling for
Modeling
4
Scientific Method
1. The Inductive method is used to establish
reliable associations among sets of facts.
e.g., edge vegetation in fields positively
correlated with an index of game abundance
• We would be using induction if we declared a
law of association; the more trials observed,
the more reliability we would attribute to the
law.
• Limitation: It does not explain the processes
that drive nature
Scientific Method
2. The Retroductive method is used to establish research
hypotheses about observed associations.
•
e.g., if we observed birds caching seeds more on south
slopes than on north slopes, and, our best guess was
that south slopes tended to be freer of snow than north
slopes, we would be using method of
retroduction to generalize a hypothesis.
•
The method of retroduction is the method
of circumstantial evidence used in courts
of law.
•
Not always reliable, because alternative hypotheses
can often be generated from the same set of facts.
Scientific Method
3. The Hypothetico-Deductive (H-D) method is a
means for testing research hypotheses.
• Complements the method of retroduction
• Experiment corroborates or rejects the predicted
facts
• Hypothesis is strengthened or rejected
• Thus, the H-D method is a way of gauging the
reliability of research hypothesis
5
Scientific Method and Experimental
Design
The statistical design of experiments
or “how to wrestle ecological data
from the real world to test
predictions of hypotheses.”
Scientific Method
# of mice trapped
Induction
There is a reliable
association
between # of mice
trapped and rainy
nights in a month
Rainy nights in a month
Obviously, there is a pattern, but, what is
the process that causes the pattern?
“Induction” cannot answer this question.
Scientific Method
# of mice trapped
To answer the question; “what is the process
that causes the pattern”, we turn to….
Rainy nights in a month
Problem?
Retroduction
Our research hypotheses
about the association
between # of mice trapped
and rainy nights is that mice
prefer rain-soaked seeds
(due to aroma).
Alternate hypotheses e.g.,
mice come out on rainy nights
to avoid predation.
6
Scientific Method
# of mice trapped
To test the question; “does moisture enhance
food preference in mice?, we turn to….
Hypothetico-deduction
We predict that mice would
take more wet seeds than dry
seeds.
Rainy nights in a month
We need an experiment!
Basic principles of Experimental Design
• Appropriate experimental design is crucial to
the acquisition of scientific knowledge.
• Hypothesis: Possible explanation for a
phenomenon. Scientific investigation
involves the testing of predictions based on
hypotheses.
• Hypothesis testing: modern science
advances by the rejection of hypotheses.
There is no such thing as proof.
Basic principles of Experimental Design
• Choose appropriate questions – good design does
not correct an inadequate grasp of the problem: e.g.,
– 1. Availability of wet seeds increases the size of a
mouse food cache
– 2. Wet seeds will benefit the overall ecology of mice
• The first hypothesis is testable against a predicted
outcome
• The second hypothesis covers everything and
probably cannot be disproved. What is “overall
ecology?”, How is it measured?
7
Basic principles of Experimental Design
• Hypothesis – Research questions are usually
phrased in a positive form. e.g., Do mice take
more wet seeds than dry seeds?
• The question is most easily tested statistically
if we frame it in negative form, Ho: There is no
difference in the number of wet and dry seeds
that mice collect. (Mice take wet and dry
seeds equally).
Scientific Method
There are 2 types of statistical hypotheses for any research
hypothesis-testing situation.
• The null hypothesis (H0) states that there is no
difference between a measured parameter and a
specific value, or between two parameters.
Null hypothesis: Ho: µ1 = µ2
• The alternative hypothesis (HA) states a specific
difference between a parameter and a specific value
or between two parameters.
Alternative hypothesis: Ha = µ1 ≠ µ2
Basic principles of Experimental
Design
• If this hypothesis is falsified by data showing more
wet seeds than dry seeds, we reject the null
hypothesis in favor of an alternative hypothesis, HA.
• Whereas a question can generate only one Ho,
there may be a number of competing HA. In the
mouse example, the alternatives to no difference in
wet or dry seeds
– Mice take fewer wet seeds
– Mice take more wet seeds
– Mice choose depressions in the ground to look for seeds
and water naturally collects there so seeds are wet
because of location.
8
Basic principles of Experimental
Design
• Replication: Let N≥ 2
• Replication is the repetition of the basic
experimental unit
• Replication reduces effects of random error, chance
events, and initial or inherent variability among
experimental units
• Pseudo-replication: The taking of sub-samples of a
single treatment and their invalid use as ‘replicates’
Basic principles of Experimental Design
• Randomization: randomize whenever possible
– Randomization means that both the allocation of
experimental units and the order in which trials are
conducted, are randomly determined.
– Proper randomization also “averages out” the effects of
extraneous factors that may be present.
– Randomization removes experimenter bias
• Control: That level of a factor subjected to zero
treatment. (not necessarily left undisturbed).
– Every ecological field experiment must have a
contemporaneous control treatment.
– Control treatments allow assessment of temporal change
and procedure effects.
Scientific Method
Hypothetico-deduction
Lab experiment:
Transparent boxes filled with
sand, covered with litter;
sunflower seeds, mice,
paper towel.
Sunflower seeds
Dry seeds
Wet seeds
Hidden wet seeds
Hidden dry seeds
Ho: There is no difference in mice preference for wet and dry seeds.
HA: There is a difference in mice preference for wet and dry seeds.
9
Research Sequence
1. Pose a research question (usually our best
guess or prediction as to what is going on)
2. Convert that to a null hypothesis
3. Design the experiment
4. Collect the data that will test the null
hypothesis
5. Run the appropriate statistical test
6. Accept or reject the null hypothesis in the light
of that testing
7. Convert the statistical conclusion to a
biological conclusion
Hypothesis testing
• There is always a chance that we are wrong
A statistical test uses the data from a sample to
make a decision about whether the null
hypothesis should be rejected or not rejected.
There are 4 possible
results when hypothesis
testing:
Hypothesis testing
An incorrect decision can occur
by one of two ways …..
A type I error occurs
when a true null is rejected
(α).
A type II error occurs
when a false null
is not rejected. (ß)
α
β
10
Hypothesis testing
• Standard statistical test concentrates on
minimizing Type I error.
• Extent of minimizing Type II errors is called
power.
• Depending on context, avoidance of Type II
may be more important than ensuring
warranted rejection of the null hypothesis
Hypothesis testing
• α is often called the significance level of the
statistical test
• In ecology we often set α = 0.05 (i.e., the
probability that an event will occur by chance
only 1/20 times)
• Usually referred to as P = 0.05
• The P = 0.05 is a gentleman’s agreement
Ten Rules for Ecological Experiments
(C.J. Krebs 1999. Ecological Methodology)
1. Not everything that can be measured should be.
–
Use ecological theory and insight to help you decide what
is interesting and important.
2. Find a problem and state your objectives clearly.
–
Find a problem that, once solved, will have many
ramifications in ecological theory or in the management
and conservation of our resources.
3. Collect data that will achieve your objectives and
make a statistician happy.
–
–
Collect enough data to reach a firm conclusion.
But make sure you understand what the statistician is
telling you.
11
4. Some ecological questions are impossible to
answer at the present time.
– Some ecologists pick exceedingly interesting but
completely impossible problems to study.
– Keep in mind the population of interest when
designing your sampling methods.
There are 3 populations in the real world:
Sample measured
Statistical
Inference
Study Population
Population of Interest
Ecological
Inference
5. Save time and money by deciding on the number
of significant figures in the data before you start
an experiment.
6. Never report an ecological estimate without some
measure of its possible error.
– Remember that every measurement no matter how
hard won, is still subject to error.
7. Be skeptical about the results of statistical tests
of significance.
– View the results of statistical tests as shades of
gray, but not black and white.
– For example, does P = 0.051 mean that an effect is
not significant?
8. Never confuse statistical significance with
biological significance.
– Biological significance is not a mechanical concept
like statistical significance. Be wary of ignoring
ecological significance.
9. Code all of your ecological data and enter it on a
computer in some standard format.
– Data management and analysis is enhanced by
organizing electronic systems of data management
(Excel, Access etc.)
10. Garbage in, garbage out.
– Always be alert to potential bias and errors
and try to eliminate them.
12