NEWS.md

In jazznbass/scan: Single-Case Data Analyses for Single and Multiple Baseline Designs

scan 0.61.999

New

Changes

• tau_u(): New method "tarlow" calculates Tau-U as implemented in an R code and online calculator by Tarlow (2017). Here, tau values are calculated as in the method = "complete", continuity_correction = TRUE, tau_method = "a". Inferential statistics are calculated based an tau b and the standard deviation for S is derived directly from Kendall's Tau B analysis (different from the parker and complete methods).
• tau_u(): Method "parker" ignores the tau_method setting and sets continuity_correction = FALSE. This follows the Parker (2011) paper. There, the inferential statistics are calculated using Kendall's Tau b while the actual Tau calculation applies Kendall's Tau a (without ties).

Solved bugs

scan 0.61.0

Solved bugs

• Corrected the name of the level-effect predictor for regressions when the phase variable is not named "phase".

New

• Tip-of-the-day like message at start-up.
• Multiple improvements of the Shiny app (try out with shinyscan())
• new output engine for rendering html export based on gt table. Set options(scan.export.engine = "gt"). This engine allows to export tables into docx format: overlap(exampleAB) |> export(file = "test.docx", flip = TRUE).
• new export functions for pem(), pet(), pnd(), and summary() (either summary(exampleAB) |> export() or export(exampleAB, summary = TRUE))

Changes

• rci(): removed the Hageman et al. method as it is not appropriate for single-cases in the current implementation.
• scdf(): New argument phase_starts(). Which defines the measurement times of the start of each phase. phase_starts = list(A = 1, B = 10, C = 15). It throws an error when a phase start is defined where no corresponding measurement-time exists. phase_starts is a generalization of B_start.
• rand_test(): New option for statistic: SMD calculates the standardized mean difference as Hedge's g with Durlak correction. W-test computes Wilcoxon tests and compares average W statistics. T-test computes T-tests and compares average t-Values. NAP and NAP decreasing for Non-overlap of all pairs.
• nap(): added Cohen's d and R-Squared effects.
• export(): select argument for nap
• coef.sc_hplm(): new casewise argument. If set TRUE, returns the effect estimations casewise.
• print.sc_hplm(): new casewise argument. If set TRUE, returns the effect estimations casewise.
• export.sc_hplm(): new casewise argument. If set TRUE, returns the effect estimations casewise.
• export.scdf(): new summary argument. If TRUE, returns a summary.
• hplm(): new arguments random_trend, random_level, and random_slope to selectively add respective random slope effects to the model.

New examples

• Parker2009b

scan 0.60.0

New function

• ird(): Robust improvement rate difference as formulated by Postejovski (2019).

Changes

• pand(): Rewrote function. New argument method allows to apply the sorting algorithm proposed in Parker 2007 and method = "minimum" applies the exact method provided by Pustejovski in 2019. Furthermore, the tau test was replaced with a X Squared and a Fisher exact test.
• corrected_tau: Report results with warning when all phase A data are identical; new argument tau_method = "a" to switch to Kendall's tau-a.
• export(): new for pand().
• New example datasets: Tarlow2017, Parker2011b, Parker2009, Parker2007

Shiny scan

• Added power-analyses
• Added settings
• Extended save options
• Various changes and optimization of the ui

scan 0.59.0

• describe(): now works correctly for duplicated phase names.
• shiny-app no longer depends on shinyjs and markdown. shinyscan() asks if missing packages scplot and shiny should be installed automatically.

scan 0.58

Shiny app added

• start the app with shinyscan()
• you need the following packages to run the app:
• shiny, shinyjs, scplot, markdown

New fucntions

• batch_apply(): Apply a function to each element in an scdf. Use . as a placeholder for the scdf case.

batch_apply(exampleAB, plm(.) |> coef())

• Helper functions for transform(): n(), all_cases(), across_cases(), first_of
• Helper functions for transform(): moving_mean(), moving_median(), local_regression()
• corrected_tau(): returns multiple cases
• export(): new for nap(), pen()
• c()/combine(): new arguments to set author and info attributes of the resulting scdf (author, info)

Changes in functions

• plm(): Print function allows to set maximum lag for autocorrelations; Overall significance Ljung_Box test is reported.

plm(exampleAB$Johanna) |> print(lag_max = 5)

Bug fixes

• solved: summary() failed when one of the cases had no name
• solved: nap() only reported values for the first case with multiple cases.
• solved: add_l2() lost column name of l2 variable when l2 had only one variable.

minor

• convert(): new arguments. indent sets the indentation. When the scdf contains only one case, no study is combined.
• select_phases(): New argument phase_names sets names of the recombined phases. The default "auto" creates combinations of the phase names automatically (e.g., A = c("A", "B", B = "C") results in phases AB and C).

scan 0.56

New features

• export(): New export for power_test() and smd() output.

• export(): tau_u() export with new argument case which takes the values "meta" or "all" and new argument select allowing to select, reorder, and rename specific variables.

• select_cases(): Allow for a selection based on object names (like in substitute).

select_cases(exampleAB, -c(Johanna, Karolina))
select_cases(exampleAB, Johanna, Karolina)
v <- c("Moritz", "Jannis")
select_cases(exampleA1B1A2B2, v)

• plm(), hplm(): New arguments contrast_level and contrast_slope allow for setting the contrasts for level and slope separately. Both elements can either be "first" or "preceding".
• Speed-up tau_u() by 20%.
• rewrote as_scdf(), read_scdf(). Now it is easier to import data from any file format.

readODS::read_ods("filename.ods") |> as.scdf()
openxlsx::read.xlsx("filename.xlsx") |> as.scdf()


readODS::read_ods("filename.ods") |> 
  as.scdf(
    cvar = "id", 
    pvar = "section", 
    mvar = "day", 
    phase_names = c("baseline", "intervention")
  )

as.data.frame(exampleABC) |> readODS::write_ods("filename.xlsx")
as.data.frame(exampleABC) |> openxlsx::write.xlsx("filename.xlsx")

Bug fixes

• tau_u(): Rewrote the calculation of meta analyses and confidence intervals.
• tau_u(), correted_tau(): corrected a wrong calculation of the continuity correction when values where lower in phase B.
• tau_u(): Implemented a new method for calculating confidence intervals based on Fisher-Z transformations (see Long, J. D., & Cliff, N. (1997). Confidence intervals for Kendall’s tau. British Journal of Mathematical and Statistical Psychology, 50(1), 31-41.

superseded function

• smooth_caes(), shift(), standardise(), ranks(), truncate_phases(): All superseded by transform() and its helper functions. See details in the help files of transform and in the scan-book.

minor changes

• as.data.frane.scdf()/as_scdf(): keep and retrieve scdf attributes.

scan 0.55

new functions

• coef()/ coefficients(): Method for base R coef function for plm/hplm objects. Extracts coefficient tables from provided object.

coef(plm(exampleAB$Johanna))

New features

• power_test(): New argument ci provides confidence intervals for power, alpha error, and correct proportions. New arguments binom_test_alpha, binom_test_power, binom_test_correct provide test against a provided proportion for alpha, power, and correct proportions.

design <- design(
  n = 1, phase_design = list(A = 6, B = 9),
  rtt = 0.8, level = 1.0, trend = 0.05
)

power_test(
  design, ci = 0.95, binom_test = TRUE
)

• plm(), hplm(): Added contrast argument with values "first" or "preceding". Now model = "JW" is deprecated and identical to model = "B&L-B", contrast = "preceding".
• plm(), hplm(): Added model "W". Which shifts the measurement-time variable to start with zero. This leads to a more sensible estimation of the intercept (where the intercept depicts the estimated score at the start). model = "W" is now the default. Use model = "B&L-B" for previous defaults.

Bug fixes

• solved #66: set_vars() working.
• read_scdf() now works when cvar is not the first column.

Changes to functions

• corrected_tau(): changed default setting of arguments to: repeated = FALSE, continuity = FALSE. Now the default results match the calculator developed by Tarlow.

• tau_u(): Implemented a continuity_correction (S-1 for calculating Z)

scan 0.54.1

Bug fixes

• select_phases() now works when phase variable is not "phase"
• estimate_design() was broken and is fixed now.
• print() function for sc_designfixed.

scan 0.54

new functions

• transform(): Takes an scdf and calculates or modifies variables for each case (transform(exampleAB, z_values = scale(values), t_values = 50 + z_values * 10)).
• smd() reporting various types of standardized mean differences.

reanmed functions (old functionnames still work)

• readSC() -> read_scdf()
• writeSC() -> write_scdf()
• design_rSC() -> design()
• rSC() -> random_scdf()

Complete rework - as new

• power_test() with various extensions, optimizations, and solved various bugs. rewrote the print method, added an argument duration to print the computation duration. Added the 'n_trials' argument for binomial distributions. Extended the help page.
• design() and its print method. Extended the help page. Rewrote the algorithm for the 'binomial' distribution.

Extended functions

• plm(): rewrote the analysis function for binomial tests. These now need an argument var_trials to define the number of trials per measurement. The dvar_percentage argument must be set TRUE when the dependent variables are percentages (and family = 'binomal').
• speed optimized random_scdf(). Rewrote the algorithm for 'poisson' distributed measures. Rewrote the algorithm for the 'binomial' distribution. Extended the help page.
• read_scdf(): extracts filetype from file extension.
• read_scdf(): New yaml import options for scdf files

Anna:
  values:
    A: [1, 3, 4, 5, 6, 7]
    B: [8, 9, 10, 10, 11]

Toni:
  values:
    A: [2, 3, 4, 5, 6, 7]
    B: [3, 9, 10, 10,11]
  control_var: [1,2,3,4,5,6,7,8,1,2,3]

• tau_u() #51: Added option for confidence intervals for tau_u output.

Solved error in functions

• describe(): solved wrong calculation of Hedges G when phase length differed.
• plm() solved #46: throws no error, when a phase is of length 1.
• corrected_tau() solved #48: throws warning when A phase has less than three rows.
• solved #49: changes class from tibble to data.frame within scdf.

scan 0.53

Major changes

• scdf files now allow to combine studies with different phase designs. Several functions have been adapted to handle cases with differing designs in a mutual analysis.
• The %>% operator has been imported and exported from the magrittr package. Now that R 4.1 has a pipe operator, pipes seem to become the standard. For compatibility with older R Versions, we will stay with the %>% operator for some time before switching to |>.
• To allow for a piping code, we added several functions: add_l2, select_phases, select_cases, subset, set_vars, set_dvar, set_mvar, set_pvar.

New functions

• sample_names(): Returns a character vector of length n with names by randomly drawing from a name list: type = {"neutral", "female", "male", "mixed"}. Useful to anonymize scdf files

names(exampleAB) <- sample_names(3)

-add_l2(): Adds the variables from a second level 2 data frame to an scdf matched by an id variable (default is case).

Leidig2018 %>%
  add_l2(Leidig2018_l2) %>%
  hplm(update.fixed = .~. + gender + migration + ITRF_TOTAL*phaseB, 
       slope = FALSE, random.slopes = FALSE, lr.test = FALSE)

• select_phases(): selects and recombines phases into A and B phase (equivalent to the phases argument of various functions, but useful when using %>% operators).

exampleA1B1A2B2 %>% 
  select_phases(A = c(1, 3), B = c(2, 4)) %>%
  overlap()

• set_vars(): change the core variables of an scdf (arguments dvar for dependent variable, pvar for phase variable, and mvar for measurement-time variable).

• set_dvar(), set_mvar(), set_pvar(): Shortcuts to set dvar, mvar, or pvar in a piping script e.g. exmpleAB_add %>% set_dvar("depression") %>% describe()

exampleAB_add %>%
  set_vars(dv = "depression") %>%
  overlap()

• is.scdf(): Tests if an object is of type "scdf" or not.
• check_scdf(): Checks for the validity of an scdf object (mainly used for internal tests)
• convert(): Creates an scdf syntax file from an scdf object.

# Create a syntax to code the scdf exampleAB and write it into an R file
convert(exampleAB, file = "cases.R")

• cdc: Applies the Conservative Dual-Criterion Method (CDC; Fisher, Kelley, & Lomas, 2003) to scdf objects.

cdc(Beretvas2008)
cdc(exampleAB_decreasing, decreasing = TRUE, trend.method = "bisplit")

Changes in functions

• overlap(): Added Hedges-g.
• new trend lines added to plot.scdf(): Koenig's bi-split / quarter intersect (lines = "trendA_bisplit") and Tukey's tri-split / Wald's slope (lines = "trendA_trisplit").

plot(exampleAB_50[8], lines = "trendA_bisplit")
plot(example_A24, lines = "trendA_trisplit")

• plot.scdf(): Now allows for multiple lines with different line styles.

plot(
  exampleAB, 
  lines = list(
    list(type = "median", col = "red", lwd = 0.5),
    list(type = "trend", col = "blue", lty = "dashed", lwd = 2),
    list(type = "loreg", f = 0.2, col = "green", lty = "solid", lwd = 1)
  )
)

• tau_u(): Solved bug in meta analysis #6. Reworked the complete function to be more clear and accurate. Added method_meta switching between fixed and random-effect meta analyses. Reworked the print function to look nicer.

• export(): Reworked the html output. Added a basic output for tau_u. Arguments caption and footnote allow to specify appearance (if left NA object specific output is generated.). booktab = TRUE is now set as a default for kable options.

Deleted deprecated functions

The following functions were deprecated since 2017 and are now removed from scan:

• makesingleSC()
• makeSCDF(): Please use scdf instead.
• estimateSC(): Please use estimate_design()
• power.testSC(): Please use power_test()

Bugs

• print.scdf() now prints cases when all variable names are wider than the current screen with.

scan 0.52

Major changes

• describe() as the new alias for describeSC()
• plot.scdf(), style_plot(): New options to style casenames: names which takes a list with tag = value structure. Example:

new_style <- style_plot()
new_style$names$side <- 3
new_style$names$line <- -1.7
new_style$names$col <- "darkred"
new_style$names$cex <- 1.5
new_style$names$at <- 20
new_style$names$adj <- 1
new_style$names$font <- 3
plot(exampleAB_decreasing, style = new_style)

• plot.scdf(), style_plot(): Different background colors for different phases:

new_style <- style_plot()
new_style$fill.bg <- c("aliceblue", "mistyrose1", "honeydew")
new_style$lty.seperators <- 0
plot(exampleABC, style = new_style)

plot(exampleABAB, style = c("default", "phase_shade"))

scan 0.50.5

Major changes

• fixed bug in plot function

scan 0.50.4

Major changes

• New subset() function (method from base subset() for selecting variables, rows, and cases. It takes the arguments subset, select, and cases.

subset(exampleAB, (values < 60 & phase == "A") | (values >= 60 & phase == "B"))
subset(exampleAB_add, select = c(-cigarrets, -depression))
subset(exampleA1B1A2B2, phase %in% c("A1", "B2"), cases = Pawel:Moritz)

scan 0.50.2

Major changes

• New select_cases() function.

select_cases(exampleAB, "Johanna", "Karolina")
select_cases(exampleAB, 1,2)
select_cases(exampleAB, "-Johanna")

scan 0.50

Major changes

• Started dropping the SC extension from function names e.g. overlapSC() becomes overlap()

scan 0.40

CRAN release 2019-08-11

scan 0.20

CRAN release 2016-10-15

jazznbass/scan documentation built on July 12, 2024, 6:02 p.m.