BIO 112
Hypothesis testing
A typical scientific study
(‘the scientific method’)
●
Making Observations (think of a question)
●
Formulate a Hypothesis
●
Design & Conduct Controlled Experiment
●
Analyze Data
●
Make inferences
●
Communicate results
Hypothesis (plural Hypotheses)
●
Proposed scientific explanation for a set of
observations
- an educated guess
●
Makes predictions that can be tested using data
●
Integrates prior knowledge and observations
●
Imagination, creativity
Observations
Questions
Hypothesis # 1: Hypothesis # 2:
Dead batteries Burnt-out bulb
Prediction: Prediction:
Replacing batteries Replacing bulb
will fix problem will fix problem
Test prediction Test prediction
Source: AP Biology Test falsifies hypothesis Test does not falsify hypothesis
Null hypothesis
●
Hypotheses cannot be proven, they can only be falsified or
rejected
- there may be a million observations that support a hypothesis, but we
cannot be sure that the next observation will not falsify the hypothesis
●
Therefore, in practice, we formulate a null hypothesis,
such that when the data reject the null hypothesis, there is
support for the alternative hypothesis.
●
The alternative hypothesis is the hypothesis of our
interest
Controlled Experiment
●
Take two (or more) groups that are similar (often, a single
group is divided into two subgroups randomly) and
manipulate a single variable. The groups (or subgroups) are
called ‘treatments’
●
Independent Variable: the variable being manipulated
●
Dependent Variable: the variable that changes as a response
to the independent variable. This is the variable that is
measured/observed, i.e. from which get data
Avoid manipulating more than one variable at a time.
Example: One wants to test the influence of latitude
on body size of a particular species. Can we measure
body sizes from these populations?
1500
●
m
800 m
300 m
0m
Sample sizes
1) In the previous experiment, would it be OK to simply
collect data from two populations?
2) If we wanted to test whether North Indians are taller than
South Indians, is it enough to measure heights of 2 North
Indians and South Indians each?
No. Because, this does not account for
variation/randomness
A scientific experiment usually relies on data from many
samples. In other words, a good study will have good sample
sizes.
In example 2 above, a sample can be the height of an
individual.
In other cases, sample sizes may be related to the number of
‘trials’
Lets take a hypothetical data set on heights of two
populations, with a good sample size. Can we
conclude whether or not one population is taller?
Population 1 Population 2 Population 1 Population 2
Statistical tests
●
Objective way to make inferences, especially
about differences
●
Statistical tests take into account
– Sample sizes
– Variation between groups (i.e. treatments)
– Variation within groups
Experimental design
●
Do plants grow taller under high or low moisture?
●
Lets formulate a hypothesis and design an
experiment to test it
Average size of individuals in a population A is higher than that
in population B.
B A
Beetle image source: icon-library.net
Exercise: Formulate specific hypotheses (at least 3) and design
experiments to test these hypothesis
