Vous êtes sur la page 1sur 21

MSB S2 Class Notes

Week 17 L1 Intro to behavioral ecology


Behavior everything an animal does and how it does it (ie. Migration,
muscular movement)
Field of behavioral ecology was organized by Niko Tinbergen.
Current Causes
-

Proximate - defining what the behavior is


We ask how does it work? How is the behavior triggered and
what physiological response does it involve?

Ultimate background reasoning for behavior


We ask how did it develop? How did the behavior arise in this
individual?

Historical Causes
-

Proximate
We ask what is it for? Why would the organism ever behave like
that?

Ultimate
We ask how did it evolve? How was this behavior shaped over
generations?

Summary:
-

behavioral ecology is the study of behavior


four questions to be answered
proximate and ultimate causes
current and historical causes
proximate causation, development, function, and evolution

Week 17 L2 Sampling Distributions


Estimating the mean: When we have a large population, we often
need to estimate the relevant statistics such as the mean, etc. by
random sampling.
-

the larger the sample the more reliable our estimation is


larger samples have greater chance of representing the whole population
accurately

Sampling error
Several small samples drawn from the same population generally
provide different values of the same statistics because the several
samples differ in the degree of how representative they are of the
whole population
Sampling Distribution: the distribution of given
statistics/observations over samples taken from a population
Central Limit Theorem (CLT): Means are normally distributed no
matter whether the original data are normally distributed.

The proof of the theorem implies:


-

the mean of means in the population mean


standard deviation of means is calculated by standard deviation divided by the
square root of the number of individuals in samples

Summary:
-

Central Limit Theorem


The distribution of the means of samples drawn randomly
from a given population is normal
The mean of means Is the population mean
The standard deviation of the means is called standard error

CLT holds regardless of whether the observations in the


original dara set is normally distributed

Week 18 L1 Development of Behavior


The development of behavior interaction of genetic and environmental factors
(experience)
Nature vs Nurture
-

Nature: the genetic component


Nurture: the environmental component

Some behaviors are the same in nearly all individuals in nearly all environments
- Strong nature component
- Innate behaviors
Some behaviors are variable depending on environmental experience
- Strong nurture component

Strong Genetic Components


Taxis: automatic/oriented movement
- Positive taxis: toward a stimulus
- Negative taxis: away from a stimulus
Kinesis: a change in activity level in response to a stimulus
Chemical Communication: communication through pheromones
Fixed Action Pattern: a behavioral pattern (i.e. routine), which occurs the same way each
time
- Triggered by a sign stimulus or releaser
Imprinting: environmental influence occurs only during a sensitive or critical period
(childhood) (e.g. geese imprint on their mother) has some environmental components
Mate choice: the decision for a mate some environmental influence (varies)
Habituation: reduced reaction in response to a repeated stimulus (the adapting to new
stimuli) (e.g. dogs overtime may stop barking at the door bell)

Week 18 L2 Confidence Intervals I


-

CLT states that the means of samples are normally distributed


The mean of the means is the population mean
The standard deviation of the means is a standard error (S.E.)

The probability that the mean of a given sample falls under the area between mu 1.96
S.E. covers is 95% (the probability that the population mean falls within 1.96 S.E. of
the mean of a sample is 95%)
The probability that the mean of a given sample falls under the area between mu 2.58
S.E. covers is 99% (the probability that the population mean falls within 2.58 S.E. of
the mean of a sample is 99%)
-

We are 95% confident that a population mean falls within


1.96 S.E. of the mean of a sample

We are 99% confident that a population mean falls within


2.58 S.E. of the mean of a sample

The intervals given by the critical values are called confidence intervals:
(critical value)*(S.E.)
Standard Error = standard deviation/(sqrt of # of
observations)
For samples with under 30 observations we use t score

Week 19 L1 Learning
What is learned behavior?
-

Any time behavior changes as a result of experience, it can be considered learning

It is not already established at birth how (if at all) the organism will respond to a
given stimulus

Minimal cases of learning: imprinting, habituation

Associative Learning:
-

Associating one aspect of the environment with another and modifying behavior
accordingly

Classical Conditioning:
-

Linking a behavioral response to a new, arbitrary, or neutral stimulus

Operant Conditioning:
-

A behavior is associated with a reward or punishment (trial and error learning)

Insight:
-

Problem-solving
Ability to reason, seemingly in advance or trial and error
Applying past experiences to a novel situation

Social Learning:
-

Learning from other individuals (observing/imitating or guided exploration)


As behaviors become more stable and widespread it can be seen as establishment
of culture

Week 19 L2 Data Transformation

Data Transformation: any systematic change made to the values in a data set
Data Transformation Techniques:
Data Presentation:
-

Data is sometimes difficult to interpret when it is presented as raw data


After transformation, the general tendency of the data can be more clearly seen

Normalization:
-

Many methods are based on the assumption that data is normally distributed (ie.
Parametric methods)

Many distributions are actually skewed

Count data often exhibit binomial, Poisson, or negative binomial distributions,


and these can be noticeably skewed

To apply certain (parametric) statistical methods, we may need to normalize the


data by data transformation

Variance Stabilization:
-

Some statistical methods compare two or more samples, but are based on the
assumption that the variances of the samples are similar

Data transformation can help to ensure hat this assumption is satisfied,


stabilizing the variance

Commonly used data transformations:


-

Logarithmic transformation
Square root transformation
Arcsine transformation

Back Transformation: If we transform data in order to use parametric techniques,


we need to transform the results back into the original scale

Week 19 L3 Confidence Intervals II

As long as sample sizes are large, confidence intervals for count data are calculated
exactly as they are for continuous data.
When sample sizes are small, confidence intervals for count data require an additional
correction.
Variance larger than mean logarithmic transformation
Small/Large Sample and Continuous/Discontinuous Data:

Week 20 L1 Evolution and Natural Selection


Darwinian Evolution:
Evolution (change in a species over time):
-

Contrasting view: fixed species with an unchanging essence


Minimal evolution: fixed species types, but with changes (eg. improvement)
over time

Two aspects of Evolution:


Macroevolution: Change in observable traits or phenotypes
Microevolution: Change in underlying gene frequencies or genotypes

Common Ancestry:

Additional evolutionary idea: the branching tree of life (phylogeny)


Part of learning about a traits evolution is tracing its emergence in a phylogenetic
tree

Establishing phylogenetic relationships:


-

Shared characteristics (homologous structures, genetic similarities)

Natural Selection:
- A mechanism that can account for evolution (not the only possible mechanism)

Evolution of novel, complex traits:


-

Gradual process
No foresight! Needs to be advantageous (or at least neutral) at ever step
Limited by available options (recall: homology)

Evolution of new species


-

Generally thought to involve some factor e.g. a geographic barrier leading to


reproductive isolation between two groups, then evolution occurring in each
group separately.

Three types of selection:


-

Directional selection: an extreme phenotype is favored over other


phenotypes
Disruptive selection: the variance of the trait increases and the population
is divided into two distinct groups
Stabilizing selection: genetic diversity decreases and the population mean
stabilizes on a particular trait value

Week 20 L2 Statistical Testing:


Null hypothesis (H0):
- Usually, it is the chance hypothesis, ie. An explanation of the observed data in
terms of sampling error
-

Usually, it is the hypothesis that there is not a trend or pattern of a certain kind in
the populationthough there might appear to be one based on the sample, or we
initially thought there could be one and wanted to test it

Dependably, it is the hypothesis for which we can mathematically model the


distribution of results allowing us to calculate a precise probability for our
particular result

Possible outcomes of statistical tests:


Null hypothesis is true and accepted
Null hypothesis is true and rejected type 1 error
Null hypothesis is false and accepted type 2 error
Null hypothesis is false and rejected statistical power
As the risk of type 1 error goes up the risk of type 2 error goes
down (and vice versa)
We can precisely control the risk of type 1 error:
- 95% significance level means that we are accepting a 5% chance of type 1 error
- 99% significance level means that we are accepting a 1% chance of type 1 error
The risk of type 2 error depends on:
- The significance level
- The amount of sampling error
- The extent of the actual population-level trend or pattern
(standard error variance within the data and sample size)
Statistical power:
The chance of correctly identifying a real population-level trend or pattern (i.e. rejecting
null hypothesis when it is false)
If there is a high chance of making a type 2 error when statistical power is low (and vice
versa)

Summary:

Week 22 L1 Comparing Averages 1: Parametric Tests


Parametric tests for comparing averages are
more restrictive than non-parametric tests (i.e. more assumptions need to
be satisfied)
generally considered more powerful/dependable

Main assumptions of parametric methods:


Data is measured on interval or ratio scale
Data is approximately normal (Can be checked in usual ways such as
shape of normal distribution, etc. and can transform certain data e.g.
counts of things and proportions to produce normal distribution.)
NOTE: what really matters is the shape of the sampling distribution, so as
long as the samples are large, the distribution of the original data isnt a
concern.

The sampling distribution for mean differences:


Is a normal distribution

Is centered on the actual mean difference, namely (mean1-mean2)


According to null hypothesis, it is centered on 0
Standard Error For Mean Difference Formula:

The formula does not assume anything about the variances being similar to each
other
Z-Score/Test Summary

For a sample mean difference the Z/T score is calculated by:

Standard Error and Z-score Formulas for Sample Mean Difference (simplified):

Week 22 L2 Comparing Averages II: Parametric Tests


Standard Error for mean difference is larger than a single mean.
When we need to do a t-score (at least one sample size is <30) we take the smaller
of the two sample sizes and multiply by 1.5 or you can use an online calculator to
find the degree of freedom for the t test.
Week 23 L1 The Work of Gregor Mendel, Part I
Gregor Mendel:
Austrian Monk
Studied math and science at university
Conducted systematic experiments of inheritance in his abbey garden
Genes: genetic determinants
Alleles: variations of the gene
Sexually reproducing organisms inherit one allele fro a characteristic from each
parent.
Dominant Alleles: Automatically determine the trait
Recessive Alleles: Can hide
Week 23 L2 Comparing Averages III
There are both parametric and non parametric tests for comparing averages
Parametric Tests:

For observations on interval or ratio scales only


Compare means
Statistical hypotheses
For all methods, the relevant hypotheses are:
H0: The samples are taken from populations with the same average
Accept provisionally if the result is not statistically significant)
H1: The samples are taken from populations with different averages
Accept provisionally if the result is statistically significant
Our choice of test depends on
a. the number of samples being compared
b. the kind of samples being compared
The Mann-Whitney U-test compares two unmatched samples
The Kruskal-Wallis Test compares more than two unmatched samples
The Wilcoxon test compares two matched samples
Matched and Unmatched Samples
When we collect data from the same individuals at two different times, for
example, the resulting observations are uniquely linked and are said to be
matched.
When observations are not linked in this way they are called unmatched
Mann-Whitney U-Test: nonparametric technique for comparing the
medians of two unmatched samples.
May be used for samples with as few as 4 observations
Does not assume that the underlying distribution is normal or that
variances are similar
Is applied on ranks, to which interval and ratio scale observations must be
converted

Week 23 L2 Comparing Averages IV


The Kruskal-Wallis test: simple non-parametric test for comparing the medians
of three or more samples
-

Observations can be on ordinal or interval/ratio scale (interval/ratio is converted


to ordinal)

Underlying distributions do not have to be normal and variances do not have to be


similar

Samples do not have to be of equal size

IF there are just 3 samples, there should be more than five observations in each
sample

The test statistic K is evaluated using the x2-distribution table

Wilcoxon Test: Simple non-parametric test for comparing the medians of two
matched samples
-

The test statistic T is evaluated using the probability distribution given in


Appendix 7

Unlike other non-parametric tests, it is applicable only if observations are on


interval/ratio scale

The null hypothesis is rejected if T is less than or equal to the threshold and
accepted if T is greater than the threshold.

Week 24 L1 The Work of Gregor Mendel II


Parental generation is true-breeding
- when they self-fertilize, they always produce offspring with the same traits
- Both copies of the gene must be the same

*A capital letter represents a dominant allele: a lowercase represents a recessive allele


A homozygous trait is made up of two of the same alleles
A heterozygous trait is made up of two different alleles
Mendels Law of Independent Assortment
- The inheritance of one trait is independent of the inheritance of other traits
Week 25 L1 - Statistics Review
Sampling Distribution
-

The distribution of a given statistic for multiple random samples taken from a
single population

Central Limit Theorem


-

The distribution of sample means (as well as various other sample statistics) is
normal
The mean of the sample means is the population mean
The standard deviation of the sample means is called the standard error
and is calculated by:
mean/sqrt of # of observations
The CLT holds whether or not the observations in the original data set(s)
are normally distributed

Confidence Intervals for the Mean:


-

Sample mean (critical value) x (S.E.)

Why we use data transformation:


- Data presentation
- Normalization
- Variance stabilization
Common transformations
- Logarithmic count data, variance>mean
- Square root count data, variance roughly = mean
- Arcsine proportion data
Four Possible Results of Statistical Tests:

Comparing Averages:
Parametric Methods:

Assumptions
Interval ratio scale data
Normally distributed data
Populations with similar variances
Compare Means
F test Z test or T test (or confidence interval)

Non-Parametric Methods
-

Assumptions
Ordinal/interval/ratio scale data
Compare Medians
Mann-Whitney U-Test
Kruskal-Wallis Test
Wilcoxon Test

F test
- Calculate variance ratio
- Variances are similar enough if F value is below critical value
Z and T test
- Null Hypothesis: samples were dreawn from populations with a mean difference
of 0
- Accept null hypothesis if t or z score is below critical value, reject if above
- Critical values and standard error formulas depend on sample size
-

Confidence intervals
(sample mean difference) (critical value) x (S.E.)

The Mann-Whitney U-Test


- Compares medians of two unmatched samples
- Accept null hypothesis if test result is above critical value, reject if below
The Kruskal-Wallis Test
- Compares medians of three or more unmatched samples
- Accept null hypothesis if test result is below critical value, reject if above
The Wilcoxon Test
- Compares medians of two matched samples
- Accept null hypothesis if test result is above critical value, reject if below
Week 25 L2 Biology Review

Tinbergens 4 causes
-

Mechanism/Triggering Cuase
How does it work? How is the behavior triggered and what physiological
response does it involve?

Development (ontogeny)
How does it develop? How does the behavior arise in an individual?

Function
What is it for? What advantages does the behavior provide for the
organism?

Evolution (phylogeny)
How did it evolve? How was this behavior shaped over multiple
generations?

Development of Behavior:
-

Nature vs Nurture
Nature: the genetic component
Nurture: the environmental component

Some behaviors are the same in nearly all individuals, in nearly all environments
Strong nature component
Innate behaviors

Some behaviors are variable, depending on environmental experience


Strong nurture component

Genes and Environment:


-

Innate Behaviors
Taxis
Kinesis
Fixed Action Patters
Chemical Communications (pheromones)

Environmental Influence
Example: mate choice
Learning
Learning:
-

Habituation

Imprinting
Sensitive Period

Associative Learning
Classical Conditioning
Operant Conditioning

Insight
Social/Observational Learning
Culture

Evolution of Behavior:
-

Genes affect behavior, therefore behavior can evolve through natural selection

Survival
Defense or Predator Avoidance
Foraging

Reproduction
Mate Selection
Competition for Mates

Evolution: 3 Ideas:
-

Evolution (change in population over time)


Common Ancestry
Natural Selection

Inheritance Vocabulary
-

True Breeding: if you cross two true-breeding individuals with the same trait, all
offspring will have that trait.
Alleles: variations of a gene
Genotype: the combination of alleles that an individual has
Homozygous: both alleles are the same

Heterozygous: the two alleles are different

Phenotype: the trait that results from the expression of the genotype

Dominant Allele: always affect the phenotype

Recessive Allele: only affects the phenotype when a dominant allele is not present

Mendels Law of Inheritance:


- Law of Segregation
Two alleles for the same gene
One passed on from parent to offspring
-

Law of Independent Assortment


The inheritance of two genes will occur independently

Vous aimerez peut-être aussi