spgarbet
commented
4 years ago That N value in the hmisc transform shows how many "Endpoints at close-out" were defined. What you're wanting is the column N value instead. If there were multiple rows of different variables this would end up being the same value over and over and not be descriptive of that rows data. One can see this by adding another row of data to the table that has a differing number of NAs present.
library(tangram)
pbc$surv <- factor(sample(c("Yes", "No", NA), length(pbc$drug), replace=TRUE))
sum(!is.na(pbc$surv))
tangram(drug ~ surv + sex, pbc, collapse_single=FALSE)Which gives the following
> tangram(drug ~ surv + sex, pbc, collapse_single=FALSE)
===============================================================
N D-penicillamine placebo not randomized
(N=154) (N=158) (N=106)
---------------------------------------------------------------
surv 269
No 0.495 45/91 0.457 48/105 0.479 35/73
Yes 0.505 46/91 0.543 57/105 0.521 38/73
sex 418
male 0.097 15/154 0.133 21/158 0.075 8/106
female 0.903 139/154 0.867 137/158 0.925 98/106
===============================================================The recommended method for the hmisc transform to see the column summary is to use the option overall=TRUE.
> tangram(drug ~ surv + sex, pbc, collapse_single=FALSE, overall=TRUE)
===============================================================================
N D-penicillamine placebo not randomized Overall
(N=154) (N=158) (N=106) (N=418)
-------------------------------------------------------------------------------
surv 269
No 0.495 45/91 0.457 48/105 0.479 35/73 0.476 128/269
Yes 0.505 46/91 0.543 57/105 0.521 38/73 0.524 141/269
sex 418
male 0.097 15/154 0.133 21/158 0.075 8/106 0.105 44/418
female 0.903 139/154 0.867 137/158 0.925 98/106 0.895 374/418
===============================================================================If one wants to change the default summary, this is a Categorical X Categorical variable, dispatched via the "hmisc" transform, lines 446-448 at the end of transform-hmisc.R. This results in the function summarize_chisq getting called. The line at 314
table <- add_col(table, cell_style[['n']](sum(!is.na(row$data)), possible=length(row$data), ...))Is the one you'd want to change. To override this tranform, create your own transform in the current environment by copying the existing. Then define whatever you'd like.
hmisc <- tangram::hmisc
hmisc[['Categorical']][['Categorical']] <- your_function_hereThen the new behavior would be used for every call.
However, the original 'N' are descriptive of the data in a consistent manner that handles multiple rows. The dplyr approach is probably the simpler, because now you're giving a description of data that has been processed to drop those with missing columns variables.
bogie
So I encountered another problem/style question while using tangram. Currently the N column only removes NA values from datar, however it would be nice to also exclude entries which don't exist in the column.
For example I have the following table:
While there it is correct that 122 patients had an endpoint at the end of the study only 108 patients had an endpoint AND were in either in the Non-Diabetic or Diabetic group:
While I could filter for those patients using dplyr, it would probably make sense to include this in the transform itself. However I can't wrap my head around parts of the API, any help would be appreciated!