In HaoLi111/dichotomousKey: Dichotomous Key Manipulation

knitr::opts_chunk$set(echo = TRUE)

Installation

install.packages("devtools")
require(devtools)
install_github("HaoLi111/dichotomousKey")

require(dichotomousKey)#load the package

Dichotomous Key Used for classfication

Using a dk table, one can get the name of the species by answering a series of questions. For example, to classify myself as a human.

• First, I ask: am I animal-like or plant like? I am animal-like, so I go to the next question to see what kind of animal I am:
• Do I live on land or water? My answer is : on land. go to next question to classify land animals:
• Do I have feathers, or just skins and hair? Well certainly I am not an angel, so I do not have feathers. I belong to the section "mammal"

Now considering converting the series of questions to a table below:

knitr::kable(dk_eg)

Note that it follows the format of:

(words in the parenthesis () are notations, not to be included)

id(index) | P(Phenotype) | G(Genotypical reference) | ref(reference) | ... ...

Note that phenotypes and genotypes are only analogies to the entries in the dataset.

Clearly, one can start from the father node ID 1, starting from line 1,2

dk_eg[dk_eg$id==1,]

since they both have father node ID 1. I am animal, so I look at the second row, 4th column. The number there points to the next father node ID I should look for, in this case, 3.

dk_eg[dk_eg$id==3,]

Since I live on land, I go to "land animal". The reference point to node 4

dk_eg[dk_eg$id==4,]

When the reference column(4th column) is 0. I know that this table cannot classify further.

dk_classify

The process above can be automated using dk_classify(). Hence,

> me = dk_classify(dk_eg)
classification start from dk_eg 
 in interactive mode 
 to abort type'q'
Phenotypical characteristic:
Key ::  1  :: 
plants like :: || ::plants
Key ::  2  :: 
animal like :: || ::animal
2
Phenotypical characteristic:
Key ::  1  :: 
live under water :: || ::fish
Key ::  2  :: 
live on land :: || ::land_animal
2
Phenotypical characteristic:
Key ::  1  :: 
hair :: || ::mammal
Key ::  2  :: 
feathers :: || ::bird
1
> me
[1] "mammal"

For each step of interaction, the function prints the options in the console. Press 1 or 2 to answer the questions to continue.

Why use dk table?

Efficient: to classify among 2^n species, it needs n questions at its best. Deterministic: given a list of phenotypes, one cannot get 2 different answers

Functionalities

So far we know how to classify a species using a naive example. Of course you can import your table as dataframes with StringsAsFactors==F. But that is uninteresting. The text below gives a thorough instruction on the capability of the data structure.

Table-List convertions

You can convert a table form dk dataset to its list form, & vice versa. The corresponding functions are dk_as_list() and list_as_dk

dk_as_list(dk_eg)

Using other packages, put it in a better form:

library(Hmisc)
list.tree(dk_as_list(dk_eg),fill = "     -")

dk_extract: Given a specimen, extract expected phenotypes

Suppose I know I am a mammal, but I want to know what properties I have, I can extract my expected phenotypes from a dk table.

me = "mammal"
dk_extract(dk_eg,me)

dk_append() what if I want to combine 2 tables together?

dk_bird

dk_bird is a finer table that only classify with the assumption that the specimen is a bird.

dk_append(dk_eg,dk_bird,"bird")

dk_partition() Given a large dk set, partition a smaller dk table only classifying some subspecies or sub subclasses

In this case, want the smaller dk starting from questions on animals only. But since it is best to start with father node index==1, we use dk_reduce to correct the id and ref data.

dk_partition(dk_append(dk_eg,dk_bird,"bird"),"animal")
dk_reduce(dk_partition(dk_append(dk_eg,dk_bird,"bird"),"animal"))

dk_is_a(): check inheritance: does one specie belong to a given class?

dk_is_a("Small Bird","land_animal",dk_partition(dk_append(dk_eg,dk_bird,"bird"),"animal"))
dk_is_a("bird","animal",dk_eg)
dk_is_a("fish","land_animal",dk_eg)

Application on larger datasets?

Suppose I have a list of observations:

fieldwork_findings = data.frame(n = 1:10,sps=c("bird","bird","Big Bird","Small Bird","fish","mammal","fish","bird","Small Bird","Bird"))

Let's do bird count

is_bird = sapply(fieldwork_findings$sps, dk_is_a,b = "bird",dk = dk_append(dk_eg,dk_bird,"bird"))
is_bird
length(is_bird[is_bird==TRUE])

More importantly one might want to calculate diversity up to a certain level, for example, how many species does the dataset contain can differ from how many subspecies the dataset contains. However this is only sensible for large amount of data.

Prospective: Abstraction by row index: for datasets more than 4 columns

What if one wants to add links as a 5th column of the table? or maybe pictures? A sensible way to do such manipulations is to specify the output row IDs.

HaoLi111/dichotomousKey documentation built on March 25, 2021, 12:32 p.m.