Changelog • ecotox

ecotox v1.4.4

CRAN release: 2021-10-27

LC_ and LT_ functions now determine both the standard error similar to how predict() works as well as the covariance of the model to be used by ratio_test()
ratio_test() was discovered to be inaccurately calculating test statistic and p value when log_x = FALSE. This has been corrected when determining t by taking the log of each doses
citation for the package has been updated as the paper the package was writing for has now been published
v1.4.4 will be pushed to CRAN within a few days

LC_ and LT_ functions now have warnings messages for not supplying a variable for the weights argument. This is due to only needing to supply if you are taking the response / total for your response variable within the formula call of LC_probit. Otherwise if you use cbind(response, non-response) method you do not need to supply weights. If you do the model will be incorrect. If you don’t supply weights there is a warning that will help you to make sure you are using one method or the other.
This stackExchange question explains the differences to using cbind() vs. respone / total method. cbind() function in R for a logistic regression

Test coverage is now 100% for all functions including ratio_test()
LC_ and LT_ functions now have warnings for not supplying a value for p
LC_ and LT_ functions now have error messages for not supplying a variable for the weights argument.
ecotox no longer relys on ggplot2 as ggplot2 was a dependency as the examples for LC_ and LT_ functions use ggplot2 to plot the results. Instead these examples have now been commented. To run this part of the example just uncomment the lines. This was done as ggplot2 does not need to be a dependency for ecotox. tibble and stats are now the only dependency with stats being the only truely necessary dependency for ecotox. tibble is being used instead of R’s default dataframe as tibbles are more useful but not necssary for the anlysis.

type is now null with default being "probit" can be "logit" if specified
compare argument has been added. This adds a compare column to the output letting the user know what models are being compared. Current default stats "Model 1 - Model 2". However, you can specificy this with a vector of characters like compare = c("May - June")

New function, ratio_test. This function allows the user to caculate differences between two seperate LC or LT values using a ratio test derived from Wheeler et al. 2006. To use the function, the user needs to create two glm() objects with either a probit or logit link. Model objects then can be passed to ratio_test() along with a percent argument which is the same argument as the p argument in LC_ and LT_ functions. A type argument is also needed to specify the link function used in the glm(). This can be either "probit" or "logit".
All functions now output a tibble instead of dataframe
New argument log_base for all LC, LT, and ratio_test functions. It allows the user to specifcy the base used when log transfoming the dose/x variable.
All functions now allow for an LC or LT value to be calculated down to the 1.0e-16 of a percent however four decmial places is typically appropriate (e.g. LC75.5). Tibble output will usually round the p value to the nearest whole number but if you insepect the p column or export to excel using openxlsx package you’ll see p in the output is the same value supplied in the function call.

All functions now have long_output argument which allows the user to decide whether or not a short or long output is displayed. Default is TRUE which will display long output of all 19 variables. If FALSE is used only 7 variables will be displayed.

All functions now have log_x argument which allows the user to specify if the x variable was log10 transformed or not. Default is TRUE which will output back transformed results. If FALSE results will not be back transformed.
LT_logit was producing error as variance co-variance matrix was only being created if the analysis required a heterogenitity correction factor. This has been fixed and a variance co-variance matrix is produced whether or not a hetrogenity correction factor is needed.
Tests have been updated to be more comprehensive for ‘ecotox’

All functions now have subset argument
Data argument can now be just the dataframe of interest no need to use brackets to subset
Weights is no longer an optional argument and can be specified just by object name. Again no need to add the name of dataframe, bracket to subset [] and $ to select object used for weights.
LC_probit and LC_logit and LT_probit and LT_logit might be depreciated as lc_probit and lc_logit and lt_probit and lt_logit are simpler function calls for users and are consistent with naming conventions. Thoughts??

ecotox has now been up in operation for about 3 months. I’ve made some recent improvements, e.g. subset feature, and am still working on a spearman_karber function which I hope to implement within the next month or so. I have three other functions I want to implement:
1. Ratio comparison test for comparing LCs and LTs following Wheeler et al. 2006
1. A parallelism test following Robertson et al. 2007
1. A equality test following Robertson et al. 2007.

What features or functions do people want to see? What do people think of current functions and those that are currently under development?

Probit and logit function name conventions have been modified. To call a probit analysis use LC_probit or LT_probit. For logit use LC_logit or LC_probit.
Agruments within the functions and objects within the function have been renamed based on new naming convention. het.seg is now het_sig and conf.level is now conf_level.
Added tests to package

Logit function for LC and LT has been added and can be called upon using LClogit or LTlogit
LC and LT orignal functions have been renamed to LCprobit and LTprobit and LC and LT function have been depreciated and eventually will be removed from the package.

‘ecotox’ v1.1.0 is now available on CRAN
het.sig is a new argument added to the both LC and LT functions it allows for the user to adjust the level of signficance that is used to decided whether or not a hetrogenity factor is used when calculating fiducial condfidnece limits. Default is 0.15 which is the same default used by SPSS 24 when doing a probit analysis.
‘ecotox’ v1.2.0.9000 DESCRIPTION file has been updated so the package now depends on R (>= 3.3.2) instead of R (>= 3.3.3) as this lead to users who were using OS X Mavericks unable to use ‘ecotox’ as OS X Mavericks can only run R v3.3.2 and less.