Data Analysis Flashcards
Coverage
= Number and % of relationship for which the behaviour occurs (or does not occur)
Dominance hierarchies:
are networks of dominance relationships, which are
determined by asymmetrical displays of threat and, especially, subordination
Advantages of having a high rank
-Priority of Access models
- Dominant individuals get (first) priority over resources
- Can be
* Fertile females ~ reproductive success of males
* Resources, e.g food ~ reproductive success of females
* Good quality territories
Animals show different behaviour patterns depending on:
season, habitat quality, time of day, weather conditions, disturbance, age, sex
Directional consistency index:
- Calculated across all dyads
- Reflects frequency with which behaviour occurred in its most frequent direction,
relative to the total of times the behaviour occurred.
Based on matrix, three aspects can be easily measured
- Coverage
- Directional consistency
- Linearity
What is the time budget?
The time budget is a breakdown of how an animal spends its day and serves to quantify which behaviours (e.g. resting, foraging, moving, etc.) are displayed at what time and how often.
Why to investigate time budget?
The time budget provides insights into fundamental ethology and behavioural ecology and may be related to environmental changes. This allows us to see how e.g., disturbance, climatic changes, or reintroduction may affect behaviour
standard error:
S.E. =(standard deviation 𝑠 )/√(sample size n )
Sequential data
- refers to (observational)
data that has a temporal
order. - In contrast: Non-
sequential data (or non-
sequential analyses, see
e.g. time budget analysis)
can explain how animals
distribute their time
across various types of
behaviour. - But: They are unable to
show how types of
behaviour interact, e.g.,
how one type of
behaviour may trigger
another, or how types of
behaviour may be related.
Sequential analysis
is a method to describe the temporal structure of behaviour based on sequentially recorded
- Using sequential analysis we can quantify if transitions between given types of behaviours are probable, improbable, or unrelated
Important assumptions / requirements for a principal component analysis:
- PCA assumes a correlation between variables!
- PCA is sensitive to the scale of the variables!
- Always standardize or normalize variables before a PCA!
- All variables must be numeric (interval/ratio) and there must be no missing values!
- PCA is not very robust against outliers!
how to measure the number of inconsistencies?
is the number of 1’s under the diagonal
What is a Inconsistencie?
are couples of individuals whose interactions go against the order of the rank
What is the strength of inconsistencies
is the total difference in rank for each inconsistency
Landau’s h index
- Indicates how linear a dominance
hierarchy is - Between 0 and 1
Steepness
- Indicates the strength of the differences
between ranks - Between 0 (an egalitarian hierarchy) and 1
(a despotic hierarchy) - Can be visualized by making a scatterplot
with rank order against the David Score
We can also visualize our results with a Dominance graph (with arrows, looks likes an octagon):
This shows the direction of the interactions between each individual with each of the others
There are 3 cluster analysis methods in SPSS:
- Hierarchical cluster analysis
- K-Means cluster analysis
- Two-step cluster analysis
Hierarchical cluster analysis (HCA)
Is usually used when the number of objects (cases) is not large (n<100). All variables you want to use for clustering must be at the interval or ratio level.
Cluster Analysis
is a method to classify or group objects into clusters based on their characteristics or attributes
The goal is to maximise the similarities within a cluster and minimise the similarities between the clusters.
K-Means cluster analysis
You indicate the number of clusters beforehand. This method is very suitable if you have a large number of objects (cases). All variables you want to use for clustering must be at the interval or ratio level.
Two-step cluster analysis
Can cluster obejcts (cases) based on nominal or ordinal scaled variables (catagorical) as well as interval or ratio scaled variables (metric). Useful for large datasets.
Divisive clustering
All objects begin within one cluster and are removed one by one until each cluster only contains one object.
There are several methods to determine the distance between two clusters:
- Single Linkage = nearest neighbor
method - Complete Linkage = furthest neighbor
method - Average Linkage = unweighted pairgroup
method with arithmetic mean - Centroid Linkage
- Ward’s Linkage
Agglomerative Clustering
Start with a separate cluster for each object
Determine the distance between each of the clusters
Combine the most similar clusters into one
Repeat steps 2 and 3 till all objects are in 1 cluster
