Scribd
Upload
49 views

Repeatability

This document discusses measuring repeatability in animal behavior. Repeatability quantifies how consistent an individual animal's behavior is and ranges from 0 to 1. The document uses example data from measuring 15 individuals twice to calculate repeatability through analysis of variance (ANOVA). The ANOVA results show the differences between individuals are statistically significant, indicating the behaviors are repeatable. Calculating components of variance from the ANOVA output gives a repeatability value (r) of 0.526, meaning 52% of behavioral variation is due to differences between individuals.

Uploaded by

Supriya Manoj
Copyright
© © All Rights Reserved
Available Formats
Download as DOC, PDF, TXT or read online on Scribd
49 views

Repeatability

This document discusses measuring repeatability in animal behavior. Repeatability quantifies how consistent an individual animal's behavior is and ranges from 0 to 1. The document uses example data from measuring 15 individuals twice to calculate repeatability through analysis of variance (ANOVA). The ANOVA results show the differences between individuals are statistically significant, indicating the behaviors are repeatable. Calculating components of variance from the ANOVA output gives a repeatability value (r) of 0.526, meaning 52% of behavioral variation is due to differences between individuals.

Uploaded by

Supriya Manoj
Copyright
© © All Rights Reserved
Available Formats
Download as DOC, PDF, TXT or read online on Scribd
You are on page 1/ 6

Page 1 of 6

Repeatability Behavioral Ecology Fall 2001

Repeatability assesses how consistent something is; in the case of behavior we are
usually interested in how consistent an individual is (or we might be interested in our
measurement error, which can also be assessed through repeatability).

We all know that behavior is very variable, both across different individuals as well as
within an individual. Repeatability, in our behavioral context here, is a measure of how
consistent the individual is.

Why is individual consistency in behavior of interest to us?

If individuals are highly consistent, but there is a lot of difference among different
individuals, then that suggests that there may be a genetic basis to the differences in
behavior among individuals. That is the type of genetic variation that selection acts on.
On the other hand, if individuals are as variable in their own behavior as the whole
population of individuals (i.e. individuals are NOT consistent) then that suggest that any
differences between individuals may have causes other than genetic differences.

Repeatability, often symbolized r, ranges from 0 to 1 and expresses the proportion of


variation in a trait that is due to differences among individuals not due to differences
within an individual. Think for a second: if the average individual is consistent, then the
average within individual variation will be low. This will make the ratio of among
individual variation to within individual variation (the repeatability) be high.

Repeatability is calculated from Analysis of Variance (ANOVA). If you have had


statistics, more power to you, but this exercise assumes little stats background. Just go
along with the idea that ANOVA can in fact partition variation into parts due to variation
within individuals and parts due to variation among individuals.

Example, you measure something 2 times for each of 15 individuals. Because


individuals do not behave exactly the same way all the time, you notice that the
individual scores vary from the first measurement to the second. Nonetheless, it looks as
though some individuals got lower scores both times, and other individuals got higher
scores both times. Is there a significant repeatability to this behavior?
Page 2 of 6
Some pretend data:

Individual Measurement Score


1 1 0.34325629
1 2 0.14305564
2 1 0.80843765
2 2 0.97922126
3 1 0.70119901
3 2 1.23151964
4 1 1.15389224
4 2 0.88256177
5 1 1.26174179
5 2 1.52169388
6 1 0.00791876
6 2 0.85306927
7 1 1.84605942
7 2 2.03122889
8 1 2.11409613
8 2 2.09923619
9 1 2.23405768
9 2 1.65710812
10 1 1.78042119
10 2 1.32637704
11 1 0.40223491
11 2 1.46224999
12 1 1.93175326
12 2 1.14711549
13 1 0.62974894
13 2 2.49582839
14 1 2.54894476
14 2 1.92692013
15 1 2.61080975
15 2 2.0333514

This is how you might keep your data in Excel if you are a well organized individual.
But to do an ANOVA in MS Excel, you need to arrange your data like this:

Sum of Score Individual                            


Measurement 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
1 0.343 0.808 0.701 1.154 1.262 0.008 1.846 2.114 2.234 1.78 0.402 1.932 0.63 2.549 2.611
2 0.143 0.979 1.232 0.883 1.522 0.853 2.031 2.099 1.657 1.326 1.462 1.147 2.496 1.927 2.033

Experienced MS Excel users might want to use the Pivot table command to change their
data around so that each individual is a column; otherwise do it the old slow way, cut and
paste.
Page 3 of 6
Once the data are arranged so that each individual is a column, highlight the data (not the
column heading, etc, just the data) and then find Data Analysis in the Tools Menu. If it
isn’t there, select Add Ins from the Tools Menu, and then check the Analysis Tools
Pack button.
Page 4 of 6

Once you have found Data Analysis you want Anova: Single Factor

Hit ok, and this should appear:


Page 5 of 6

You have to check the Output options Output range button, and then enter a cell
destination where there is room on your worksheet, if you are unsure where there is
room, just check the New worksheet ply button. Hit ok, and this appears:
Page 6 of 6

Check that the computer cooperated. In the top part you should have the same number of
groups as you have individuals (each individual is treated as a group in Excel lingo), each
individual (group) should have a count of 2 observations (you observed each individual
twice), as well as a sum, an average, and a variance.

At the bottom is the ANOVA table. It is the information in here that you use to calculate
repeatability. Notice first that Excel gives you a statistical test of whether or not a
significant amount of the total variation is found by differences Between the different
groups (individuals) RELATIVE to the amount of variation Within the groups. The
statistic is the F test, with degrees of freedom (df) and the P-value. The P-value is the
likelihood of finding as big a difference between groups as you did find IF THERE
WERE NO TRUE DIFFERENCE.

Here F with 14 and 15 df = 3.217, P = 0.016. That means that you have a 1.6% chance
that there is NOT a true difference between groups. If this chance is low enough, then we
might provisionally say that the groups ARE different. Usually in most of science if the
chance is 5% or less we say it is statistically significant.

Ok, so there is a 1.6% chance that the groups (individuals) are NOT different, so it looks
as though they behaviors are in fact repeatable. So…….

How do we calculate the repeatability r?

r = S2A/(S2 + S2A)

Where S2A is the between group variance and S2 is the within group variance.

The S2 = the mean square within (MSW). Excel gives the MSW as the Within row MS
column = 0.257 in this example.

S2A = (MSA - MSW)/n0

And no is the weighted average number of observation per group (here we have 2
observations per individual, so no = 2, but you could have the case where you have
different numbers of observations in each group. In which case:

N0 = [1/(a-1)] * [SUM(ni) – (SUM(ni2)/SUM(ni))

So, in this example:

S2 = MSW = 0.257

S2A = (MSA - MSW)/n0 = (0.828 – 0.257)/2 = 0.2855

r = S2A/(S2 + S2A) = 0.2855/(0.257 + 0.2855) = 0.526

That means that about 52% of the variation is due to differences among individuals,
which can only happen if individuals are consistent. Recall that r ranges from 0 to 1, so
an r of 0.5 is pretty consistent. Good Stuff Maynard.

You might also like

pFad - Phonifier reborn

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.

Note: This service is not intended for secure transactions such as banking, social media, email, or purchasing. Use at your own risk. We assume no liability whatsoever for broken pages.


Alternative Proxies:

Alternative Proxy

pFad Proxy

pFad v3 Proxy

pFad v4 Proxy