momx / momocs Goto Github PK

• add an option for plotting only on PC axis
• plot mshapes at the origin when calculation pos.shp
• more rigorous nc.shp and nr.shp

• for error / fit visualization ala Lipo et al.
• or directly onto the mean shape
• or linearly (ala cal_q)

reported by John.

all is said

between sets of measurements :

• with hclust and monophyletic clades
• with hpow and others

• allow to define a pipe for subsetting views, doing calibration, etc. then combining them

a big baby

and overall, a less hideous and more generic morphospace

• make subset accept a column name and return a list with split of all levels
• make combine accept such a list

• should look like Terral 2010

Try to convert existing plots into ggplot2 paradigm. At least, return them or bridge existing methods

all is said

• should make something better than: calibrate_harmonicpower(bot, "efourier")$minh
• print.listMomocs + invisible

to easily plot sample sizes from a Coo/Coe/PCA, etc. object. anything with a $fac.

not sure though

to be used for multivariate regression, or as another graphic layer in PCA/LDA plotters

Turn all graphs into ggplot2 graphs.

So far (0.99) only the core functions are available. Integrate them all in Opn.

Got the code but have to integrate it for various methods.

• subsetting with ids : slice
• subsetting with factor name : chop
• subsetting with logical : filter

all is said

eg hpow for efourier, retain for LDA on PCA, etc.

From pixels to cm/mm ; when scale is different or not but available for every shape.

very confusing so far

around core functions already present (methods on Opn, morphospace, etc.)

• should be based on a maximal fit fir the best shape.

update this subset version so that it will accept sth like "type" rather than "type=='bot'" and then return a list of Out objects

• should be working with arg "range", not "harm.range"
• calibrate_deviation plotter should be adding lines for %s of relative deviation

• type shoudl be "u" by default
• integrate pvclust
• intergrate shapes

• set a wrapper to retain 99% (or x%) of PC/LD

• While R is not the best place to do so, and before an ImageJ plugin, continuous acquisition eg landmark definition / scale acquisition
• Also, a better handling of errors (eg only failed imports would have to be redone) for import_jpg and import_txt (and possibly others)

maybe a better way to print those objects (eg shapes[4])

• implement MANOVA formula to allow 2+ factors.

• map $ to fac$ in Coo and Coe classes (eg bot$type <=> bot$fac$type)

when n is higher than the number of points, should either shift to interpolate, with a message; or return an error.

a simple wrapper.

and propose it when var-cov are not equal

• allow it to return a list of all levels of a given fac column

eg: > rf$confusion
beer whisky class.error
beer 16 4 0.20
whisky 5 15 0.25

• coo_slidedirection should allow to not center shapes (eg ig they have been Procrustes / baselined before).
• then, it should return id0-s to allow to change ldk accordingly
• coo_slide.Coo should allow different options : a single id1, a list of id1s (eg from coo_slidedirection), or a single ldk. ldk, if any, should be changed accordingly.