Package 'allestimates'

Estimates all possible effect estimates using Cox Proportional Hazards regression models

Description

Estimates hazard ratios using Proportional Hazards Regression models ("coxph" from survival package) from models with all possible combinations of a list of variables.

Usage

all_cox(crude, xlist, data, na_omit = TRUE, ...)

Arguments

crude	An object of formula for initial model, generally crude model. However, any other variables can also be included here as the initial model. The left-hand side of ~ is the outcome of interest, and the variable on the right-hand side of ~ is the exposure of the interest (either a treatment or a risk factor)
xlist	A vector of a list of variable names.
data	Data frame.
na_omit	Remove all missing values. Default is "na_omit = TRUE".
...	Further optional arguments.

Value

A list of all effect estimates.

See Also

surival

Examples

vlist <- c("Age", "Sex", "Married", "BMI", "Education", "Income")
results <- all_cox(crude = "Surv(t0, t1, Endpoint) ~ Diabetes", xlist = vlist, data = diab_df)
results

---

Estimates all possible effect estimates using glm

Description

all_glm estimates odds ratios or rate ratios using generalized linear models (glm) with all possible combinations of a list of variables (potential confounding factors).

Usage

all_glm(crude, xlist, data, family = "binomial", na_omit = TRUE, ...)

Arguments

crude	An object of formula for initial model, generally crude model. However, any other variables can also be included here as the initial model.
xlist	A vector of a list of variable names (potential confounding factors).
data	Data frame.
family	family Description of the error distribution. Default is "binomial".
na_omit	Remove all missing values. Default is "na_omit = TRUE".
...	Further optional arguments.

Value

A list of all effect estimates.

See Also

stats

Examples

diab_df$Overweight <- as.numeric(diab_df$BMI >= 25)
vlist <- c("Age", "Sex", "Income")
all_glm(crude = "Diabetes ~ Overweight", xlist = vlist, data = diab_df)

---

Estimates all possible effect estimates using lm

Description

all_lm estimates coefficients of a specific variable using linear models (lm) with all possible combinations of other variables (potential confounding factors).

Usage

all_lm(crude, xlist, data, na_omit = TRUE, ...)

Arguments

crude	An object of formula for initial model, generally crude model. However, additional variables can also be included here as the initial model.
xlist	A vector of a list of variable names (potential confounding factors).
data	Data frame.
na_omit	Remove all missing values. Default is "na_omit = TRUE".
...	Further optional arguments.

Value

A list of all effect estimates.

See Also

lm

Examples

vlist <- c("Age", "Sex", "Cancer", "CVD", "Education", "Income")
all_lm(crude = "BMI ~ Married", xlist = vlist, data = diab_df)

---

Plot all effect estimates against p values

Description

all_plot() generates a scatter plot with effect estimates of all possible models again p values.

Usage

all_plot(
  data,
  xlabels = c(0, 0.001, 0.01, 0.05, 0.2, 0.5, 1),
  xlim = c(0, 1),
  xlab = "P value",
  ylim = NULL,
  ylab = NULL,
  yscale_log = FALSE,
  title = NULL
)

Arguments

data	Object from all_cox, all_glm, all_speedglm, or all_glm, including all effect estimate values.
xlabels	Numeric vector x-axis tick labels. Default is "c(0, 0.001, 0.01, 0.05, 0.2, 0.5, 1)".
xlim	Vector of 2 numeric values for x-axis limits. Default is "c(0, 1)".
xlab	Character string for x-axis name. Default is "P value".
ylim	Vector of 2 numeric values for y-axis limits.
ylab	Character string for y-axis name. Default depends on original model types.
yscale_log	TRUE or FALSE to re-scale y-axis to "log10". Default is "FALSE".
title	Character for plot title. Default is "NULL".

Value

A ggplot2 object: scatter plot

Examples

vlist <- c("Age", "Sex", "Married", "BMI", "Education", "Income")
results <- all_cox(crude = "Surv(t0, t1, Endpoint) ~ Diabetes", xlist = vlist, data = diab_df)
all_plot(results)

---

Draws scatter plot with all effect estimates against AIC

Description

all_plot_aic() generates a scatter plot with all effect estimates against AIC.

Usage

all_plot_aic(data, xlab = "AIC", ylab = NULL, title = NULL)

Arguments

data	Object from all_cox, all_glm, all_speedglm, or all_glm, including all effect estimate values.
xlab	Character string for x-axis name. Default is "AIC"
ylab	Character string for y-axis name. Default depends on original model types.
title	Character for plot title. Default is "NULL".

Value

A ggplot2 object: scatter plot

Examples

vlist <- c("Age", "Sex", "Married", "BMI", "Education", "Income")
results <- all_cox(crude = "Surv(t0, t1, Endpoint) ~ Diabetes", xlist = vlist, data = diab_df)
all_plot_aic(results)

---

Draws multiple scatter plots of all effect estimates against AIC

Description

all_plot_aic2() draws multiple scatter plots of all effect estimates against AIC. Each plot indicates if a specific variable is included in the models.

Usage

all_plot_aic2(data, xlab = "AIC", ylab = NULL, title = NULL)

Arguments

data	Object from all_cox, all_glm, all_speedglm, or all_glm, including all effect estimate values.
xlab	Character string for x-axis name. Default is "AIC".
ylab	Character string for y-axis name. Default depends on original model types.
title	Character for plot title. Default is "NULL".

Value

A ggplot2 object: scatter plot.

Examples

vlist <- c("Age", "Sex", "Married", "BMI", "Education", "Income")
results <- all_cox(crude = "Surv(t0, t1, Endpoint) ~ Diabetes", xlist = vlist, data = diab_df)
all_plot_aic(data = results)

---

Plots all effect estimates against p values with each specific variable in the models

Description

all_plot2() generates a panel of scatter plots with effect estimates of all possible models again p values. Each plot includes effect estimates from all models including a specific variable.

Usage

all_plot2(
  data,
  xlabels = c(0, 0.001, 0.01, 0.05, 0.2, 0.5, 1),
  xlim = c(0, 1),
  xlab = "P value",
  ylim = NULL,
  ylab = NULL,
  yscale_log = FALSE,
  title = NULL
)

Arguments

data	Object from all_cox, all_glm, all_speedglm, or all_glm, including all effect estimate values.
xlabels	numeric vector x-axis tick labels. Default is "c(0, 0.001, 0.01, 0.05, 0.2, 0.5, 1)"
xlim	vector of 2 numeric values for x-axis limits. Default is "c(0, 1)".
xlab	Character string for x-axis name. Default is "P value".
ylim	vector of 2 numeric values for y-axis limits.
ylab	Character string for y-axis name. Default depends on original model types.
yscale_log	TRUE or FALSE re-scale y-axis to "log10". Default is "FALSE".
title	Character title. Default is "NULL".

Value

A ggplot2 object: scatter plot

Examples

vlist <- c("Age", "Sex", "Married", "BMI", "Income")
results <- all_cox(crude = "Surv(t0, t1, Endpoint) ~ Diabetes", xlist = vlist, data = diab_df)
all_plot2(results)

---

Effect estimates from models with all possible combinations of variables

Description

To assess treatment effects in clinical trials and risk factors in bio-medical and epidemiological research, we use regression coefficients, odds ratios or hazard ratios as effect estimates. allestimates allows users to quickly obtain effect estimates from models with all possible combinations of a list of variables specified by users. all_lm for linear regression, all_glm for logistic regression, all_speedglm using speedlm as a faster alternative of all_glm, and all_cox for Cox Proportional Hazards Models. Users can further use those values in a returned list of results. all_plot draws scatter plots with all effect estimate values against p values, as Stata confall command (Wang Z (2007) <doi:10.1177/1536867X0700700203>). Those plots divide estimates into four categories:

Details

• positive and significant: left-top quarter

• negative and significant: left-bottom quarter

• positive and non-significant: right-top quarter

• negative and non-significant: right-bottom quarter

all_plot2 draws multiple plots. Each of those plots indicates whether a specific variable is included or not included in models. Those effect estimates help users better understand confounding effects, uncertainty of their estimates, as well as inappropriately including variables in the models. This is a tool for calculating and exploring effect estimates from all possible models. Interpretation of the results should be in the context of other analyses and biological knowledge.

Examples

? all_speedglm
? all_glm
? all_cox
? all_lm
? all_plot
? all_plot2

---

Example data: Health outcomes of 2372 adults with and without diabetes

Description

A data frame with 2372 rows and 14 variables with diabetes status diabetes and mortality status endpoint. For the purpose of demonstrate, assume that we are interested in the association between diabetes and endpoint. Other variables are considered as possible confounding factors. The purposes of this dataset is to illustrate those functions in chest and allestimates packages only. Therefore, we assume it is a cohort design for Cox Proportional Hazard regression, and a case-control design for logistic regression.

Usage

diab_df

Format

A data frame with 2372 rows and 14 variables:

Diabetes

diabetes status 1: with diabetes 0: without diabetes

Endpoint

mortality status 1: reached end point, and 0: survived

Age

Age, in years

Sex

sex, 1: male, 2: Female

BMI

Body mass index

Married

marital status 1: married, 0: not

Smoke

smoking status 1: smoker, 0: non-smoker

CVD

cardiovascular disease 1: yes 0: no

Cancer

cancer 1: yes, 0: no

Education

education 1: high, 0: low

Income

income 1: high, 0: low

t0

time (age) at the start of the follow-up

t1

time (age) at the end of the follow-up

mid

matched set id, for conditional logistic regression

---