merge(by=...) for geomorph_data_frame

[Bob LePaul]

Dear colleagues,

for downstream analyses I would like to merge data from a geomorph_data_frame and a normal data_frame. However, the function merge() is not applicable for this kind of data frame, so I am looking for a suitable alternative. The geomorph documentation does not mention merging processes with a "by" variable.

Specifically, I have a dataset that I run gpagen() on and merge the coordinates into a geomorph_data_frame that also contains an "id" variable. Each "id" refers to an unique animal.

Now I have another dataset that contains the same "ids", but it is a dataset where every "id" is present more than once (individuals with the same id were used multiple times). I would like to have the same coordinates and centroid sizes assigned to every instance where the same ID is present. Usually, I would use merge(Data1, Data2, by="id"), but this method does not work for geomorph_data_frame formats. Hence, I would be grateful if you could point me to a suitable alternative.

Thank you for your time!

Bob

[Adams, Dean [EEOB]]

If you first convert your 3D array of coordinates into a 2D array, this should be possible.

Dean

 

Dr. Dean C. Adams

Director of Graduate Education, EEB Program

Professor

Department of Ecology, Evolution, and Organismal Biology

Iowa State University

https://www.eeob.iastate.edu/faculty/adams/

phone: 515-294-3834

--
You received this message because you are subscribed to the Google Groups "geomorph R package" group.
To unsubscribe from this group and stop receiving emails from it, send an email to geomorph-r-pack...@googlegroups.com.
To view this discussion on the web, visit https://groups.google.com/d/msgid/geomorph-r-package/070767c8-ba91-4478-97af-893dbf19079bn%40googlegroups.com.

[Bob LePaul]

Dear Dean,

that was what I thought and did, but it did not help. Maybe I have been overlooking something?

par_id <- read.table("ids.txt", header=TRUE, stringsAsFactors=TRUE, sep="\t", na.strings="NA", dec=".", strip.white=TRUE)

id_fam <- read.table("id_fam.txt", header=TRUE, stringsAsFactors=TRUE, sep="\t", na.strings="NA", dec=".", strip.white=TRUE)

par_data <- readland.tps("data.tps",  specID = c("ID"),  negNA = FALSE,  warnmsg = TRUE)

par.shape <- gpagen(par_data)$coords
par.csize <- gpagen(par_data)$Csize
par.dat <- two.d.array(par.shape)
par.dat.full <- geomorph.data.frame(coords =par.dat, csize = par.csize, id=par_id$ID)

par.dat.full2 <- merge(par.dat.full, id_fam, by="ID", all.x=TRUE)

ERROR: cannot coerce class '"geomorph.data.frame"' to a data.frame

[Adams, Dean [EEOB]]

Taking a 2D matrix and then putting back into a geomorph data frame is unnecessary.

To view this discussion on the web, visit https://groups.google.com/d/msgid/geomorph-r-package/5b72b545-37ac-4550-86a9-60c7d0ac8a84n%40googlegroups.com.

[Mike Collyer]

There is no merge option for geomorph.data.frame.  You should not expect that geomorph.data.frame objects and data.frame objects are commensurate objects.  

However, merge is a pretty unnecessary function, anyway.  If you want to append objects to a geomorph.data.frame object, like a data.frame object, it is best to remember these are different types of organized lists.  Therefore, you can just treat them like lists.  So, if you have A containing $habitat and B containing $coords and $Csize, you can add $habitat to B, as

B$habitat <- A$habitat

Hope that helps.

Mike

[Bob LePaul]

Dear Mike,

the issue is the overlooked "by" argument of merge(). Lets say I have coords for ID1 in Dataset 1 and ID1 appears multiple times in Dataset2. The "by" argument of merge() allows me to match the IDs and insert the coords and csizes multiple times, wherever the appropriate ID is present. Your method unfortunately only allows appending lists with predefined order to each other.

Cheers

Bob

[Bob LePaul]

Without putting back the 2D matrix into a geomorph data frame, I am unsure how I would append another dataset containing the correct IDs for each coordinate in the correct order (the IDs derived from the tps file are unsuitable for the matching that I am looking for here).

[Bryan H. Juarez]

Hi all,

Perhaps the easiest solution is to do all the necessary data manipulation before using geomorph.data.frame(). One can use methods of their choosing to organize data after obtaining the two.d.array(), and then use geomorph.data.frame() as the last step. 

geomorph.data.frame() serves to format data in a consistent structure before fitting a model within geomorph/RRPP, and trying to manipulate it further may result in unintended changes to the structure, limiting its ability to perform its original task. It seems to me that a geomorph.data.frame() object should not generally be manipulated if we want it to interact with e.g. procD.pgls() in the intended manner.

Best,

Bryan 

To view this discussion on the web, visit https://groups.google.com/d/msgid/geomorph-r-package/5e652e27-3c0a-4a39-b383-9a6a33ce9cf9n%40googlegroups.com.

--

Bryan H. Juarez (he/him/his)

NSF Fellow

PhD Candidate

Adams Lab

EEOB Dept.

Iowa State University

909-733-3368

Twitter: @bhjuarez

[Mike Collyer]

Bob,

You just need to be a bit more resourceful.  

B$habitat <- A$habitat[match(B$ID1, A$ID1)]

You should get comfortable with functions like match, order, sort, and %in%.  These give you far more control over your data.  The merge function is a convenience function that does stuff like the above for you.  However, it only works with objects that could be coerced into data.frame objects.  Comparatively, vectors, matrices, arrays, and lists can all be sorted, ordered, partitioned, and appended, if one is comfortable base syntax.

Mike

To view this discussion on the web, visit https://groups.google.com/d/msgid/geomorph-r-package/99df7ea7-4d42-4ea7-bb8f-7e70b81d796an%40googlegroups.com.