RStudio::conf

• Talk recordings and workshop materials from rstudio::conf(2022)
  

• Also some talks are on youtube

• Some announcements:

  • RStudio is changing its name to Posit
  • Use Quarto for creating content with Python, R, Julia, and Observable
  • New developments in Shiny: Shiny for Python, Shiny without a server, a visual Shiny UI editor, and more
  • Updates from the tidymodels and vetiver teams
  • Other topics

RStudio::conf

• [Keynote] RStudio: 2022 and Beyond

• [Keynote] The Past and Future of Shiny

• [Talk] What they forgot to teach you about industry transitions from academia (WTF AITA)

Individual R user meeting

• Let’s start! Choose your time slot here
• Meet with An-Chi and Eva to discuss your R scripts and any R-related topics
• What you need to do:
  • Answer a short questionnaire about the usage of R tools
  • (preferrably) Prepare one or a few R scripts and explain to us; the goal is to see if there is any problem or space for improvement, either on your script or on the tools
  • Besides the scripts, any questions or suggestions about the tools are welcomed

Individual R user meeting (cont.)

• The meeting is voluntary while we encourage you to have it even if you only use a bit of the tools. It gives us a chance to understand your needs and develop suitable tools. If you’re an experienced user, we’d like to have your opinions to make better tools.

s2dv::Load or startR::Start?

• These are the two main tools in our department to load netCDF data (R-wise)
• The priority of Load() is low as long as Start() can achieve the same goal.
• Load() is still used by CST_Load(). But you can load data with Start() then transform the class to s2dv_cube by CSTools::as.s2dv_cube.
• as.s2dv_cube() is not perfect transforming startR object though.

s2dv::Load or startR::Start? (cont.)

→ Talk to me if you have a strong reason using Load(), so I can priorotize the development.

Tip: To read files individually, using easyNCDF or ncdf4 package is easier.

PloyEquiMap() bugfixes

When longitude is not continous (e.g., c(0:10, 350:360))

• The coastline was not complete (discussion)
• Misplaced figure (discussion)

DiffCorr() siginficance test

DiffCorr() significance test method changed. Three options: “two-sided”, “one-sided-higher”, and “one-sided-lower”.

     if (type == 'two-sided') {
        output$sign <- ifelse(!is.na(p.value) & p.value <= alpha/2, TRUE, FALSE)
      } else if (type == 'one-sided-higher') { 
        output$sign <- ifelse(!is.na(p.value) & p.value <= alpha & output$diff.corr > 0, TRUE, FALSE)
      } else if (type == 'one-sided-lower') { 
        output$sign <- ifelse(!is.na(p.value) & p.value <= alpha & output$diff.corr < 0, TRUE, FALSE)
      }

New functions

• CRPS: Compute the Continuous Ranked Probability Score
• CRPSS: Compute the Continuous Ranked Probability Skill Score

• Bias: Compute the Mean Bias
• BiasSS: Compute the Mean Bias Skill Score

dat_dim is added in several functions

• Functions that had dat_dim parameter argument (default dat_dim = ‘dataset’) indicating dataset dimension and now allow it to be NULL: Ano_CrossValid(), Corr(), RMS(), RMSS() and UltimateBrier().

RMS <- function(exp, obs, time_dim = 'sdate', dat_dim = 'dataset',
                comp_dim = NULL, limits = NULL, conf = TRUE, conf.lev = 0.95,
 ncores = NULL) { … }  

• Functions which dat_dim has been added (default dat_dim = NULL): RPS() and RPSS().

RPS <- function(exp, obs, time_dim = 'sdate', memb_dim = 'member', dat_dim = NULL,
                prob_thresholds = c(1/3, 2/3), indices_for_clim = NULL, Fair = FALSE,
                weights = NULL, ncores = NULL) { … }

dat_dim is added in several functions (cont.)

• New functions CRPS() and CRPSS() also have dat_dim parameter (default dat_dim = NULL).

New function ShiftLon()

Shift global longitudes of a data array. Useful for map plotting or aligning datasets.

ShiftLon <- function(data, lon,  westB, lon_dim = 'lon', ncores = NULL) {
}

New function ShiftLon() (cont.)

Subset bugfixes

• Bugfix for the case that only one dimension left after subsetting. It returned error before.

source("https://earth.bsc.es/gitlab/es/ClimProjDiags/-/raw/master/R/Subset.R")
res <- Subset(array(1:20, dim = c(dat = 1, lat = 4, lon = 5)), along = "lat", indices = 1, drop = T)
dim(res)
#lon
#  5

• Bugfix when param indices is not a list. It returned wrong output before.

Subset(array(1:20, dim = c(dat = 1, lat = 4, lon = 5)), 
       along = c(‘lat’, ‘lon’), 
       indices = c(1, 1), 
       drop = T)

New version 2.2.0-2 installed

NEWS - Use the destination grid to decide which indices to take after interpolation. - Bugfix when Start() parameter “return_vars” is not used. - Allow netCDF files to not have calendar attributes (force it to be standard calendar)

New release on the way…

EXPECTED NEWS - Change dependency on package s2dverification to s2dv - Bugfix of as.s2dv_cube() for startR object - CST_QuantileMapping improvement - Improve flexibility of time dimensions in s2dv_cube() - CST_Calibration for forecast - CST_BiasCorrection new parameters for dimension flexibility

User presentation [Chung]