Helper Functions for Multi-Variable Logistic Regression
Source:R/h_logistic_regression.R
h_logistic_regression.Rd![[Stable]](figures/lifecycle-stable.svg)
Helper functions used in calculations for logistic regression.
Usage
h_get_interaction_vars(fit_glm)
h_interaction_coef_name(
interaction_vars,
first_var_with_level,
second_var_with_level
)
h_or_cat_interaction(
odds_ratio_var,
interaction_var,
fit_glm,
conf_level = 0.95
)
h_or_cont_interaction(
odds_ratio_var,
interaction_var,
fit_glm,
at = NULL,
conf_level = 0.95
)
h_or_interaction(
odds_ratio_var,
interaction_var,
fit_glm,
at = NULL,
conf_level = 0.95
)
h_simple_term_labels(terms, table)
h_interaction_term_labels(terms1, terms2, table, any = FALSE)
h_glm_simple_term_extract(x, fit_glm)
h_glm_interaction_extract(x, fit_glm)
h_glm_inter_term_extract(odds_ratio_var, interaction_var, fit_glm, ...)
h_logistic_simple_terms(x, fit_glm, conf_level = 0.95)
h_logistic_inter_terms(x, fit_glm, conf_level = 0.95, at = NULL)Arguments
- fit_glm
(
glm)
logistic regression model fitted bystats::glm()with "binomial" family. Limited functionality is also available for conditional logistic regression models fitted bysurvival::clogit(), currently this is used only byextract_rsp_biomarkers().- interaction_vars
(
characterof length 2)
interaction variable names.- first_var_with_level
(
characterof length 2)
the first variable name with the interaction level.- second_var_with_level
(
characterof length 2)
the second variable name with the interaction level.- odds_ratio_var
(
string)
the odds ratio variable.- interaction_var
(
string)
the interaction variable.- conf_level
(
proportion)
confidence level of the interval.- at
(
NULLornumeric)
optional values for the interaction variable. Otherwise the median is used.- terms
(
character)
simple terms.- table
(
table)
table containing numbers for terms.- terms1
(
character)
terms for first dimension (rows).- terms2
(
character)
terms for second dimension (rows).- any
(
flag)
whether any ofterm1andterm2can be fulfilled to count the number of patients. In that case they can only be scalar (strings).- x
(
stringorcharacter)
a variable or interaction term infit_glm(depending on the helper function).- ...
additional arguments for the lower level functions.
Functions
h_get_interaction_vars(): Helper function to extract interaction variable names from a fitted model assuming only one interaction term.h_interaction_coef_name(): Helper function to get the right coefficient name from the interaction variable names and the given levels. The main value here is that the order of first and second variable is checked in theinteraction_varsinput.h_or_cat_interaction(): Helper function to calculate the odds ratio estimates for the case when both the odds ratio and the interaction variable are categorical.h_or_cont_interaction(): Helper function to calculate the odds ratio estimates for the case when either the odds ratio or the interaction variable is continuous.h_or_interaction(): Helper function to calculate the odds ratio estimates in case of an interaction. This is a wrapper forh_or_cont_interaction()andh_or_cat_interaction().h_simple_term_labels(): Helper function to construct term labels from simple terms and the table of numbers of patients.h_interaction_term_labels(): Helper function to construct term labels from interaction terms and the table of numbers of patients.h_glm_simple_term_extract(): Helper function to tabulate the main effect results of a (conditional) logistic regression model.h_glm_interaction_extract(): Helper function to tabulate the interaction term results of a logistic regression model.h_glm_inter_term_extract(): Helper function to tabulate the interaction results of a logistic regression model. This basically is a wrapper forh_or_interaction()andh_glm_simple_term_extract()which puts the results in the right data frame format.h_logistic_simple_terms(): Helper function to tabulate the results including odds ratios and confidence intervals of simple terms.h_logistic_inter_terms(): Helper function to tabulate the results including odds ratios and confidence intervals of interaction terms.
Note
We don't provide a function for the case when both variables are continuous because this does not arise in this table, as the treatment arm variable will always be involved and categorical.
Examples
library(dplyr)
library(broom)
adrs_f <- tern_ex_adrs %>%
filter(PARAMCD == "BESRSPI") %>%
filter(RACE %in% c("ASIAN", "WHITE", "BLACK OR AFRICAN AMERICAN")) %>%
mutate(
Response = case_when(AVALC %in% c("PR", "CR") ~ 1, TRUE ~ 0),
RACE = factor(RACE),
SEX = factor(SEX)
)
formatters::var_labels(adrs_f) <- c(formatters::var_labels(tern_ex_adrs), Response = "Response")
mod1 <- fit_logistic(
data = adrs_f,
variables = list(
response = "Response",
arm = "ARMCD",
covariates = c("AGE", "RACE")
)
)
#> Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred
mod2 <- fit_logistic(
data = adrs_f,
variables = list(
response = "Response",
arm = "ARMCD",
covariates = c("AGE", "RACE"),
interaction = "AGE"
)
)
#> Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred
h_glm_simple_term_extract("AGE", mod1)
#> variable variable_label term term_label interaction interaction_label
#> 1 AGE Age AGE Age
#> reference reference_label estimate std_error df pvalue
#> 1 0.1698216 0.09524116 1 0.07457501
#> is_variable_summary is_term_summary
#> 1 FALSE TRUE
h_glm_simple_term_extract("ARMCD", mod1)
#> variable variable_label term term_label interaction
#> 1 ARMCD Planned Arm Code ARM A Reference ARM A, n = 64
#> 2 ARMCD Planned Arm Code ARM B ARM B, n = 68
#> 3 ARMCD Planned Arm Code ARM C ARM C, n = 52
#> interaction_label reference reference_label estimate std_error df pvalue
#> 1 2 0.3004308
#> 2 -1.774769 1.144405 1 0.1209443
#> 3 17.1922 3626.588 1 0.9962176
#> is_variable_summary is_term_summary
#> 1 TRUE FALSE
#> 2 FALSE TRUE
#> 3 FALSE TRUE
h_glm_interaction_extract("ARMCD:AGE", mod2)
#> variable variable_label term term_label
#> 1 ARMCD:AGE Interaction of Planned Arm Code * Age ARM A Reference ARM A, n = 64
#> 2 ARMCD:AGE Interaction of Planned Arm Code * Age ARM B ARM B, n = 68
#> 3 ARMCD:AGE Interaction of Planned Arm Code * Age ARM C ARM C, n = 52
#> interaction interaction_label reference reference_label estimate std_error
#> 1
#> 2 0.3081205 0.2062392
#> 3 0.02948826 548.5923
#> df pvalue is_variable_summary is_term_summary
#> 1 2 0.3275837 TRUE FALSE
#> 2 1 0.1351767 FALSE TRUE
#> 3 1 0.9999571 FALSE TRUE
h_glm_inter_term_extract("AGE", "ARMCD", mod2)
#> variable variable_label term term_label interaction interaction_label
#> 1 AGE Age AGE Age
#> 2 AGE Age AGE Age ARMCD Planned Arm Code
#> 3 AGE Age AGE Age ARMCD Planned Arm Code
#> 4 AGE Age AGE Age ARMCD Planned Arm Code
#> reference reference_label estimate std_error odds_ratio lcl ucl
#> 1 -0.03873898 0.1514322 NA NA NA
#> 2 ARM A ARM A NA NA 0.9620018 0.7149514 1.294420
#> 3 ARM B ARM B NA NA 1.3091545 1.0021802 1.710157
#> 4 ARM C ARM C NA NA 0.9907919 0.0000000 Inf
#> df pvalue is_variable_summary is_term_summary is_reference_summary
#> 1 1 0.798092 FALSE TRUE FALSE
#> 2 NA NA FALSE FALSE TRUE
#> 3 NA NA FALSE FALSE TRUE
#> 4 NA NA FALSE FALSE TRUE
h_logistic_simple_terms("AGE", mod1)
#> variable variable_label term term_label interaction interaction_label
#> 1 AGE Age AGE Age
#> reference reference_label estimate std_error df pvalue
#> 1 0.1698216 0.09524116 1 0.07457501
#> is_variable_summary is_term_summary odds_ratio lcl ucl
#> 1 FALSE TRUE 1.185093 0.9832935 1.428308
#> ci
#> 1 0.9832935, 1.4283084
h_logistic_inter_terms(c("RACE", "AGE", "ARMCD", "AGE:ARMCD"), mod2)
#> variable variable_label term
#> 1 RACE Race ASIAN
#> 2 RACE Race BLACK OR AFRICAN AMERICAN
#> 3 RACE Race WHITE
#> 13 ARMCD Planned Arm Code ARM A
#> 23 ARMCD Planned Arm Code ARM B
#> ARM B ARMCD Planned Arm Code ARM B
#> 33 ARMCD Planned Arm Code ARM C
#> ARM C ARMCD Planned Arm Code ARM C
#> 11 AGE Age AGE
#> 21 AGE Age AGE
#> 31 AGE Age AGE
#> 4 AGE Age AGE
#> 12 AGE:ARMCD Interaction of Planned Arm Code * Age ARM A
#> 22 AGE:ARMCD Interaction of Planned Arm Code * Age ARM B
#> 32 AGE:ARMCD Interaction of Planned Arm Code * Age ARM C
#> term_label interaction interaction_label reference
#> 1 Reference ASIAN, n = 110
#> 2 BLACK OR AFRICAN AMERICAN, n = 40
#> 3 WHITE, n = 34
#> 13 Reference ARM A, n = 64
#> 23 ARM B, n = 68
#> ARM B ARM B, n = 68 AGE Age 35
#> 33 ARM C, n = 52
#> ARM C ARM C, n = 52 AGE Age 35
#> 11 Age
#> 21 Age ARMCD Planned Arm Code ARM A
#> 31 Age ARMCD Planned Arm Code ARM B
#> 4 Age ARMCD Planned Arm Code ARM C
#> 12 Reference ARM A, n = 64
#> 22 ARM B, n = 68
#> 32 ARM C, n = 52
#> reference_label estimate std_error df pvalue odds_ratio lcl
#> 1 2 0.9361139
#> 2 18.15018 3944.701 1 0.9963288 76299564 0
#> 3 -0.3727152 1.025808 1 0.7163522 0.6888614 0.09224926
#> 13 2 0.253914 NA NA
#> 23 -11.52715 6.96188 1 0.09777185 NA NA
#> ARM B 35 NA NA NA NA 0.4757148 0.03361147
#> 33 16.33327 20278.39 1 0.9993573 NA NA
#> ARM C 35 NA NA NA NA 34808099 0
#> 11 -0.03873898 0.1514322 1 0.798092 NA NA
#> 21 ARM A NA NA NA NA 0.9620018 0.7149514
#> 31 ARM B NA NA NA NA 1.309155 1.00218
#> 4 ARM C NA NA NA NA 0.9907919 0
#> 12 NA NA NA NA
#> 22 0.3081205 0.2062392 1 0.1351767 NA NA
#> 32 0.02948826 548.5923 1 0.9999571 NA NA
#> ucl is_variable_summary is_term_summary is_reference_summary
#> 1 TRUE FALSE FALSE
#> 2 Inf FALSE TRUE FALSE
#> 3 5.143998 FALSE TRUE FALSE
#> 13 NA TRUE FALSE FALSE
#> 23 NA FALSE TRUE FALSE
#> ARM B 6.732956 FALSE FALSE TRUE
#> 33 NA FALSE TRUE FALSE
#> ARM C Inf FALSE FALSE TRUE
#> 11 NA FALSE TRUE FALSE
#> 21 1.29442 FALSE FALSE TRUE
#> 31 1.710157 FALSE FALSE TRUE
#> 4 Inf FALSE FALSE TRUE
#> 12 NA TRUE FALSE FALSE
#> 22 NA FALSE TRUE FALSE
#> 32 NA FALSE TRUE FALSE
#> ci
#> 1
#> 2 0, Inf
#> 3 0.09224926, 5.14399810
#> 13 NA, NA
#> 23 NA, NA
#> ARM B 0.03361147, 6.73295611
#> 33 NA, NA
#> ARM C 0, Inf
#> 11 NA, NA
#> 21 0.7149514, 1.2944200
#> 31 1.002180, 1.710157
#> 4 0, Inf
#> 12 NA, NA
#> 22 NA, NA
#> 32 NA, NA