A Git repository was created to host some data files, in various formats. As a first step, you will have to make a local copy of this repository on your computer, and turn it into an Rstudio project.
Download the whole content of the repo as a zip file, and unzip it in a given folder on your computer.
Open Rstudio, and open the menu File > New Project.
Choose the type “Existing directory”, and create a project in the folder created in step 1.
2 Usual data formats in morphometrics
Usually, the data that we handle in morphometrics are not natively in a classic “tidy” format (one row per individual, one column per variable). They can also sometimes be in specific file formats (.dta, .xml, …).
2.1 Landmarks coordinates in text files
In the best and simplest case, the software you used to place 2D or 3D landmarks on your specimens will produce a plain text file (.csv or .txt) for each individual, with one row per landmark and one column per coordinate (x, y and optionally z) each. The basic R function read.table() will then be convenient to load your data, as in the example below.
## Exemple d'importation d'un fichier txt :colb <-read.table(file ="./scallops/colb01.txt",header =FALSE, # pas d'intitulés de colonnessep =",", # séparateur de colonnesdec ="."# séparateur décimal)head(colb) # visualiser un extrait du fichier
However, some data acquisition software will return non-standard file formats. A (non-exhaustive) list of examples is given below; the corresponding files can be explored in the formats folder of the Git repository.
2.2.1 DTA files
They are created (for instance) by the software Landmark. The data for all your individuals is generally stored in one single file. The function read.lmdta() from the R package {Morpho} can load this kind of files:
## Load a dta file:dta <- Morpho::read.lmdta(file ="./formats/example_landmark.dta")## Structure of the R object created:str(dta)
Polygon files are (for instance) created by Avizo. They may represent either landmarks data, or surfaces. They can be loaded using the function read.ply() from the package {geomorph}, or using the function vcgImport() from the package {Rvcg}.
## Load a surface in PLY format and visualize it:ply <- Rvcg::vcgImport(file ="./formats/example_avizo.ply", clean =TRUE)rgl::shade3d(ply, col ="grey")
2.2.3 TPS files
They are created by the software tpsDIG2. Once again, the data for several individuals are generally stored in one single file. The function readland.tps() from the pakage {geomorph} (or the function readallTPS() from the package {Morpho}) allows you to load such files:
## Load a TPS file:tps <- geomorph::readland.tps(file ="./formats/example_tpsdig.TPS",specID ="imageID")
No curves detected; all points appear to be fixed landmarks.
## Display the coordinates of the first shape:print(tps[, , 1])
They are created for instance by the software Viewbox. The function read_viewbox() from the package {anthrostat} allows you to load such files:
## Load an XML file:xml <- anthrostat::read_viewbox("./formats/example_viewbox.xml")head(xml)
x y z
0 39.3450 -324.3970 -23.4495
1 47.4408 -207.4721 -64.0599
2 39.8136 -169.3606 -18.9716
3 47.9278 24.9335 -173.3149
4 -48.6539 -245.8677 -19.3262
5 121.4894 -233.5452 -7.6302
There are many other classical file formats in morphometrics (e.g., JSON, NTS, …). You can find specific function to load them in the R packages {geomorph} and {Morpho}.
3 Load several files at once
For DTA or TPS formats for instance, we generally have several individuals in one single file, so that the corresponding R functions import all the individuals in one single instruction. In other cases, however, we have one file (CSV, TXT, etc.) per individual, so that we have to load a large set of data files. Obviously, if we have 80 files, we must not write 80 read.table() instructions manually!
There are 5 TXT files in this folder. How to load them with one single R instruction?
For CSV or TXT files, the function Morpho::read.csv.folder()1 loads all the files from a given directory at once, and returns an array. Of course, all files must have the same number of rows and columns, and have the same extension.
## Load all TXT files:lmarray2 <- Morpho::read.csv.folder(folder ="./scallops/", # the folder to loadx =1:44, # the rows to ready =1:3, # the columns to readheader =FALSE, # no column names in the filesdec =".", # decimal pointsep =",", # field separatorpattern ="*.txt"# pattern of the files to load)## Print the array:print(lmarray2$arr)
4 In practice
In this course, we will use the data contained in the file rats.TPS. In this file are stored the coordinates of 15 fixed landmarks (these are the first 15 rows of each individual) and 88 semilandmarks taken on rats mandibles, as shown on Figure 1.
Figure 1: Protocol for landmarks and semilandmarks.
Exercise
(Optional) Open the file rats.TPS using a text editor, inspect it and try to understand its structure.
Load this TPS file in R.
How many individuals do we have in this file?
Solution
In this TPS file, the rows LM=103 suggest that we have 103 landmarks per individual. Each landmark has two coordinates \((x,y)\), so that we have an array of 103 rows and 2 columns per individual. The file ID indicates the identifier of each individual.
Finally, we have some metadata (sex, age, weight, …) about the individuals. These data are stored in a CSV file, which is the recommended format for loading “usual” flat data in R. This data can be loaded using the built-in read.csv2() function:
Species Location Season
Rattus rattus:159 Dry forest :38 Dry:67
Field : 6 Wet:92
Hygrophilous forest:51
Mesophilic forest :49
Swamp forest :15
Weight Body_length Tail_length Age
Min. : 31.00 Min. :112.0 Min. :143.0 Min. : 37.10
1st Qu.: 62.00 1st Qu.:145.0 1st Qu.:191.5 1st Qu.: 87.95
Median : 98.00 Median :160.0 Median :210.0 Median : 225.60
Mean : 94.25 Mean :161.3 Mean :204.7 Mean : 227.04
3rd Qu.:120.00 3rd Qu.:176.5 3rd Qu.:220.0 3rd Qu.: 306.10
Max. :173.00 Max. :205.0 Max. :255.0 Max. :1015.60
Age_class Sex
Adult :88 F:92
Juvenile : 5 M:67
Older adult: 4
Sub-adult :59
NA's : 3
One useful thing, for subsequent statistical analyses, may be to convert the still unordered factor Age_class into an ordered factor. By default, the levels (i.e., age classes) have no specific order and appear in alphabetical order in all plots, which is clearly not convenient here.
## Convert age casses into an ordered factor:meta$Age_class <-factor( meta$Age_class,ordered =TRUE,levels =c("Juvenile", "Sub-adult", "Adult", "Older adult"))
As a (very strongly recommended!) good practice, we should make sure that this dataframe includes the same number of individuals as the TPS file, and that they are given in the same order.
## Check dimension of CSV file:dim(meta)
[1] 159 9
## Check consistency of ordering between TPS and CSV:all(dimnames(rats)[[3]] ==rownames(rats))
