Mantel Test - PowerPoint PPT Presentation

1 / 18
About This Presentation
Title:

Mantel Test

Description:

Evaluates the results from repeated randomizations ... Quadrat. 40 x 40 cm. Sampling Design. Data Matrices. Seed Bank. Horn-Morisita. Percent Cover ... – PowerPoint PPT presentation

Number of Views:443
Avg rating:3.0/5.0
Slides: 19
Provided by: biolo6
Category:
Tags: mantel | quadrat | test

less

Transcript and Presenter's Notes

Title: Mantel Test


1
Mantel Test
  • Evaluates correlation between distance,
    similarity, correlation or dissimilarity matrices
  • Null no relationship between matrices
  • Pearson correlation (r) can be used to measure
    strength of relationship
  • r ranges from -1 to 1

2
Test for Significance
  • Evaluates the results from repeated
    randomizations
  • If randomizations frequently produce a
    correlation stronger or as strong as the original
    data little evidence that correlation differs
    from zero.
  • Order of randomization is important
  • - order of row and columns are shuffled for one
    matrix

3
Randomization cont.
  • Mantels Z statistic is computed after each
    permutations and a distribution is created
  • Z statistic from nonrandomized data is compared
    to the distribution of the shuffled matrix.
  • If null is true, Mantel statistic will fall near
    the middle of the reference distribution
  • 10,000 randomizations is recommended
  • wij connectivity or Euclidean distance matrix
    and xij dissimilarity or distance matrix

n
Z
wijxij
n
i 1 j1 ij ji
4
What is really computed?
  • Is the relationship between distance measures not
    the raw data

5
When to use
  • Examples
  • Two groups of organisms form same set of sample
    units
  • Community structure before and after disturbance
  • Genetic distance and geographic distance
  • Ecological distance and geographic distance

6
Partial Mantel Test
  • Quantifies the relationship between two matrices
    while controlling for the effects of a third one.

7
Problem with Mantel
  • Relationship is a global outcome for all
    variables
  • Can pick out which variables have the most
    influence
  • Solution
  • Use ordination CCA and RDA

8
Vegetation Types
  • Reference Rosemary Scrub

Semi-improved Pasture
Disturbed Scrub
9
Sampling Design
Macroplot
Subplot
Quadrat
16 X 16 m
2 x 2 m
40 x 40 cm
10
Data Matrices
Seed Bank Horn-Morisita
Percent Cover Bray-Curtis
X-Y Coordinates Euclidean
11
Mantel Test in R
  • ecodist and vegan package
  • Both matrix must be identical and contain x and y
    coordinates
  • Must choose an appropriate distance measure
  • Bray-Curtis - Manhattan
  • Euclidean
  • Morisita
  • Horn-Morisita

12
Results
13
Pasture
14
Disturbed Scrub
15
Rosemary Scrub
16
(No Transcript)
17
(No Transcript)
18
Results
Write a Comment
User Comments (0)
About PowerShow.com