Skip to content

Commit

Permalink
added argument to chose sample to solve in csUTh
Browse files Browse the repository at this point in the history 
changed the example data file for csUTh
  • Loading branch information
@tonydoss
tonydoss committed Jan 13, 2019
1 parent 777f72e commit a398f42
Show file tree
Hide file tree
Showing 13 changed files with 1,387 additions and 52 deletions.
60 changes: 30 additions & 30 deletions R/functions.R
View file
Expand Up @@ -584,26 +584,28 @@ th230_u238_ratio_plot <- function(output,
#' csUTh calculates closed-system Th-230/U ages, including detrital correction.
#'
#'
#' @param input_data Input data frame, containing only rows of data for the sample that you want to solve. The following columns need to be present in this data frame: U234_U238_CORR, U234_U238_CORR_Int2SE, Th230_U238_CORR, Th230_U238_CORR_Int2SE, Th232_U238_CORR, Th232_U238_CORR_Int2SE.
#' @param input_data Input data frame, containing only rows of data for the sample that you want to solve. The following columns need to be present in this data frame: Sample_ID, U234_U238_CORR, U234_U238_CORR_Int2SE, Th230_U238_CORR, Th230_U238_CORR_Int2SE, Th232_U238_CORR, Th232_U238_CORR_Int2SE.
#' @param sample_name Name of the sample to calculate closed-system ages for. The string entered must match characters for the chosen sample in the column 'Sample_ID' of the data file. Default: 'MK16'.

This comment has been minimized.

Sign in to view

Copy link
@benmarwick

benmarwick Jan 13, 2019

Collaborator

I've been having a bit of a think about this approach here and in the other function.

I think most users would find it highly prescriptive, which we need to be thoughtful about, because ideally functions are designed for maximum flexibility to work in situations we can't imagine. Many users might find it irritating that we are so fussy about column names, etc. For example, when we use ggplot2 we don't have to change our column names to x_axis_data and y_axis_data. We tell ggplot the column names with x = my_column1 and y = my_column2, and it handles it. This the norm with functions in R, and I think we should follow this approach also.

Perhaps it's because I'm not familiar with the sources of these data for our package that I'm wondering about this. Are they all coming out of an instrument in a standard format that we can assume to be exactly the same for everyone? If they are, then it's fine to assume that everyone has the same column names, and sample IDs are formatted in the same way. If our users will have data with different column names and different sample ID formats, then we might consider having the user tell the function which columns contain what variable, and subset the specific sample on their own. That second option is more typical for R packages, and what I'd prefer. But if you tell me that everyone has the same input files coming out of their instrument, I'm fine with that too.

Is it possible that a user might have their sample IDs formatted in a different way to what you have? If so, I don't think it makes sense for us to try to imagine all the ways a sample ID might appear, and deal with them here. Instead, let's leave that to the user.

What you've got here makes sense only if we can be sure that every user will have their input data with the same column names as you. Is the input file here is coming out of an instrument or other software, so we can assume it's the same for everyone? What if that instrument or software gets an update and changes how it outputs its data?

It might be more flexible to have arguments to the function like sample_name_colum = xx and have the user tell the function what the name is of the column that has the sample ID (in that example it's xx). Currently we hard-code it as Sample_ID, and previously we hard-coded it as X. That will obviously cause errors if the user supplies a data frame with a different column name. In any case, we're forcing the user to rename the columns of their data frame, and some will find this a bit invasive because it is not common for pkg functions to require the user to change their data like that. More typical would be to get the user to tell the function what to do (i.e. , instead of the function telling the user what to do. I guess this is more of a convention I've observed in the culture of R programmers and users than a firm rule to follow.

So if we give more flexibility to the user, the function might look like this, where we expose the column names to the user in case their input data have different column names. Then they can match them up without altering their original data (which I think is what most users will expect, not to have to alter their data):

sUTh <- function(input_data, 
    sample_name_column = "Sample_ID", # we assume the user has already subset to one sample
    U234_U238_CORR = "U234_U238_CORR",
    U234_U238_CORR_Int2SE, = "U234_U238_CORR_Int2SE", 
   Th230_U238_CORR = "Th230_U238_CORR",
   Th230_U238_CORR_Int2SE = "Th230_U238_CORR_Int2SE", 
   Th232_U238_CORR = T"h232_U238_CORR",
   Th232_U238_CORR_Int2SE = "Th232_U238_CORR_Int2SE",
... and the rest...

What do you reckon?

This comment has been minimized.

Sign in to view

Copy link
@tonydoss

tonydoss Jan 14, 2019

Author Owner

I get your point but I don't think it would be too much to ask for the user to comply with the column names, so it's easy for the functions to find the data. The same applies to osUTh().
Maybe we can make easily available the CSV files used as examples for both osUTh() and csUTh(), so users can use them as templates and just put their data in?

Yes, the example data files for osUTh() are modified from outputs from Iolite software, which is a software to manipulate laser ablation data.
The example data .txt file for csUTh() was directly from Iolite. I modified it to a .csv and added the 'Sample_ID' column header (with the .txt file there was no column header where the sample names were, hence the 'X')
#' @param nbitchoice Number of iterations in the model. Recommended to have at least 100. Default: 100.
#' @param detcorrectionchoice Do a detrital correction? Enter TRUE for yes, or FALSE for no. Default: TRUE
#' @param R28det (232Th/238U) activity ratio of the detritus. Default is 0.8
#' @param R28det_err Error on the (232Th/238U) activity ratio of the detritus. Default is 0.08
#' @param R08det (230Th/238U) activity ratio of the detritus. Default is 1
#' @param R08det_err Error on the (230Th/238U) activity ratio of the detritus. Default is 0.05
#' @param R48det (234U/238U) activity ratio of the detritus. Default is 1
#' @param R48det_err Error on the (234U/238U) activity ratio of the detritus. Default is 0.02
#' @param R28det (232Th/238U) activity ratio of the detritus. Default: 0.8
#' @param R28det_err Error on the (232Th/238U) activity ratio of the detritus. Default: 0.08
#' @param R08det (230Th/238U) activity ratio of the detritus. Default: 1
#' @param R08det_err Error on the (230Th/238U) activity ratio of the detritus. Default: 0.05
#' @param R48det (234U/238U) activity ratio of the detritus. Default: 1
#' @param R48det_err Error on the (234U/238U) activity ratio of the detritus. Default: 0.02
#' @param keepfiltereddata Save filtered data on which an outlier test was performed? Only recommended if all analyses of a same sample are supposed to give the same age. Enter TRUE for yes, or FALSE for no. Default: FALSE
#' @param print_summary Print a summary of the output to the console? Default is TRUE
#' @param with_plots Draw plots? Default is TRUE
#' @param print_summary Print a summary of the output to the console? Default: TRUE
#' @param with_plots Draw plots? Default: TRUE
#'
#' @import deSolve ggplot2
#' @importFrom stats IQR optim sd
#'
#' @examples
#' data("iolite_export")
#' # Only solve for sample MK16
#' output <- csUTh(iolite_export[grepl('MK16', iolite_export$X), ],
#' data_file <- read.csv('data/example_data.csv')
#' # Solve for sample MK16
#' output <- csUTh(data_file,
#' sample_name = 'MK16',
#' nbitchoice = 100,
#' detcorrectionchoice = TRUE,
#' keepfiltereddata = FALSE,
Expand All @@ -612,6 +614,7 @@ th230_u238_ratio_plot <- function(output,
#' @export

csUTh <- function(input_data,
sample_name = 'MK16',
nbitchoice = 100,
detcorrectionchoice = TRUE,
R28det = 0.8,
Expand All @@ -635,22 +638,19 @@ csUTh <- function(input_data,
lowerbound <- c(2, 01.0) # lower bound values for age (log10(yr)) and initial (234U/238U)
upperbound <- c(6, 10.0) # upper bound values for age (log10(yr)) and initial (234U/238U)

# use detrital correction (Y) or not (N)
# use detrital correction (TRUE) or not (FALSE)
detcorrection <- detcorrectionchoice

# composition detritus
R28det <- 0.8
R28det_err <- 0.08
R08det <- 1
R08det_err <- 0.05
R48det <- 1
R48det_err <- 0.02

# import iolite results, we are assuming the user has already subset their iolite output
# so that it only includes rows of the sample they want to solve.
iolite_results <- input_data
# # composition detritus
# R28det <- 0.8
# R28det_err <- 0.08
# R08det <- 1
# R08det_err <- 0.05
# R48det <- 1
# R48det_err <- 0.02

data <- input_data
# create dataframe with data only for samples to solve
data <- subset(input_data, (grepl(sample_name, input_data$Sample_ID)))
# number of samples to solve
number_sampletosolve <- nrow(data)

Expand Down Expand Up @@ -750,11 +750,11 @@ csUTh <- function(input_data,
R48i2se_results[count] <- R48i_2se
}

final_results <- as.data.frame(cbind(as.character(data$X),
data[,27:36], data[,41:44],
final_results <- as.data.frame(cbind(as.character(data$Sample_ID),
data[,29:38], data[,43:46],
round(time_results/1000,3), round(time2se_results/1000,3),
round(R48i_results,3), round(R48i2se_results,3)))
colnames(final_results)[1] <- c("ID")
colnames(final_results)[1] <- c("Sample ID")
colnames(final_results)[16:19] <- c("Age (ka)", "Age 2se", "(234U/238U)i", "Ratio 2se")

remove_outliers <- function(x, na.rm = TRUE, ...) {
Expand Down Expand Up @@ -839,7 +839,7 @@ initial_234U_238U_plot <- function(output,
theme(panel.grid.major = element_blank(),
panel.grid.minor = element_blank())

ggplot(output, aes(ID, `(234U/238U)i`)) + # plot ages
ggplot(output, aes(`Sample ID`, `(234U/238U)i`)) + # plot ages
geom_errorbar(aes(ymin = (`(234U/238U)i` - `Ratio 2se`),
ymax = (`(234U/238U)i` + `Ratio 2se`)),
width=0.1) + # plot error bars
Expand Down Expand Up @@ -877,7 +877,7 @@ ages_plot <- function(output,
theme(panel.grid.major = element_blank(),
panel.grid.minor = element_blank())

ggplot(output, aes(ID, `Age (ka)`)) + # plot ages
ggplot(output, aes(`Sample ID`, `Age (ka)`)) + # plot ages
geom_errorbar(aes(ymin = (`Age (ka)` - `Age 2se`),
ymax = (`Age (ka)` + `Age 2se`)),
width=0.1) + # plot error bars
Expand Down
4 changes: 2 additions & 2 deletions README.Rmd
View file
Expand Up @@ -36,13 +36,13 @@ of the tooth or bone and return an age in thousands of years ago.

## Web application

The package includes a web application that runs in your browser where you can upload your CSV file, set the model parameters, run the model, and view the results. Choose this option is you are not familiar with R. Here's a screenshot of the web app:
The package includes a web application that runs `osTuh()` in your browser where you can upload your CSV file, set the model parameters, run the model, and view the results. Choose this option is you are not familiar with R. Here's a screenshot of the web app:

![](inst/iDADwigl-shiny-app.gif)

## Run code without downloading anything

You can run the package function in your browser by [starting a binder instance](https://mybinder.org/v2/gh/benmarwick/UThwigl/master?urlpath=rstudio). This will open RStudio in your browser, together with the contents of this GitHub repository.
You can run the package functions in your browser by [starting a binder instance](https://mybinder.org/v2/gh/benmarwick/UThwigl/master?urlpath=rstudio). This will open RStudio in your browser, together with the contents of this GitHub repository.

## Installing the UThwigl package on your computer

Expand Down
1 change: 0 additions & 1 deletion inst/extdata/input/IoliteExport_All_Integrations.txt
View file

This file was deleted.

0 comments on commit a398f42

Please sign in to comment.