milanwiedemann / suddengains Goto Github PK

could use base::abs(stats::qt((0.05 / 2), 4)) or something similar to change alpha, e.g.:

#' @param sg_crit3_qtp Numeric, significance level of the two-tailed student t-test to determine the critical value to be used for the third criterion.
#' Degrees of freedom are based on the number of available data in the three sessions preceding the gain and the three sessions following the gain.

e.g. see \link[package]{function}

As raised by a reviewer, it might be of interest to set individual cut offs for each participant. This would be similar to the modification of the third criterion used by Kelly et al 2005 and 2007 to assess stability.

This could be implemented in the R package as a modification of the first or third criterion. Computationally this approach is similar to the way the first criterion gets assessed.
Conceptually the way individual cut off scores are calculated would is closer to third criterion (stability). There could also be other useful applications of individual cut off scores calculated in a different way to Kelly et al.

Kelly, M. A. R., Cyranowski, J. M., & Frank, E. (2007). Sudden gains in interpersonal psychotherapy for depression. Behaviour Research and Therapy, 45(11), 2563–2572. https://doi.org/10.1016/j.brat.2007.07.007

Kelly, M. A. R., Roberts, J. E., & Ciesla, J. A. (2005). Sudden gains in cognitive behavioral treatment for depression: When do they occur and do they matter? Behaviour Research and Therapy, 43(6), 703–714. https://doi.org/10.1016/j.brat.2004.06.002

mode all the scripts I have for testing into a proper R package testing workflow

testing this in the development branch at the moment, seems like there are some problems with the way I use the write_ functions from haven

when using the plot_sg() function to visualise changes of scores around a sudden gain for two groups both groups need to have:

sg_crit123 = 1

when using the function to plot changes for one group that did have a sudden gain and another group that didnt have a sudden gain, this currently needs to be changed manually but overwriting sg_crit123 = 1 for all participants. the information about which group participants belong to (e.g., "group_variable") need to be stored in another variable and used in the group_var_name argument.

similar to sgdata, but not as a sample dataset. this will just be for testing special cases of sudden gains, e.g. missing data in pre or post

this is already possible for the first sg_crit1_cutoff = 0 and second criterion sg_crit2_cutoff = 0 but not for the third criterion. in code this could look like this: sg_crit3_cutoff = NULL. to unify the code NULL could also be added to the sg_crit2_cutoff and sg_crit2_cutoff arguments.

for example only one pre or post gain score needs to be available. there would be different options to think about adding this, for example:

• to only integrate this this only for the first and last between session interval to identify very early and very late gains/losses
• add this function for every between session interval?

Implement checks that test whether specific variables are present in the data set, e.g. "sg_session_n", "sg_magnitude" ...

set sg criteria in function, enter 3 pregain and 3 postgain values as voctor list or input (maybe also base::readline ...), this would also work great as shiny app

at the moment the option to have different colours for multiple groups doesn't work

1. check for reversal
2. check for sudden loss / sudden gain following the reversal

v0.1.0 or v0.1.1 for next release

see https://r-pkgs.org/data.html

I was thinking of something like this, get vector of all ids and then sum the duplicates and missings, this can then be checked and used for error msg

ids <- sgdata$id
sum_id_duplicates <- base::sum(base::duplicated(stats::na.omit(ids)))
sum_id_missings <- base::sum(base::is.na(ids))

for bysg datasets need to filter for unique IDs so that start values of cases with multiple gains only get selected once and not as often as that case had a sg/sl.

for byperson need to filter for sg/sl cases only so that not whole sample gets included in calculations for start and end value

this is needed when the end of tx assessment should not be part of the sg_var_list

for example class by_sg_data by_person_data, maybe this could also contain info about which gain was selected for the by_person data set

As raised during peer review, consider implementing different approaches to calculate the standard deviation for criterion 3. The method currently implemented in the package represents an independent sample t test.

• Consider the option to incorporate the correlation between the pre and post gain measures into this formula following Cohen (1988).
• Also consider the option to use only the standard deviation of the pre gain scores given that the standard deviation is maybe expected to change across the intervention Glass and colleagues (1981)

Cohen, J. (1988). Statistical Power Analysis for the Behavioral Sciences. New York, NY: Routledge Academic.
Glass, G. V., McGaw, B., and Smith, M. L. (1981). Meta-Analysis in Social Research. Beverly Hills, CA: Sage.

see this paper for the method: Hayes, A. M., Feldman, G. C., Beevers, C. G., Laurenceau, J.-P., Cardaciotto, L., & Lewis-Smith, J. (2007). Discontinuities and cognitive changes in an exposure-based cognitive therapy for depression. Journal of Consulting and Clinical Psychology, 75(3), 409–421. https://doi.org/10.1037/0022-006X.75.3.409

less copy&paste!