stmh: Calculate Incidence Rate Ratios from time-to-event data

In myominnoo/mStats_beta: A tool for data management and statistical analysis

Description

stmh calculates incidence rate ratios (IRR) of the cohort.

Usage

stmh(
  data = NULL,
  time,
  status,
  by,
  fail = 1,
  ref_value = NULL,
  strata = NULL,
  per = 1,
  rnd = 4,
  plot = TRUE,
  print.table = TRUE
)

## S3 method for class 'character'
stmh(
  data = NULL,
  time,
  status,
  by,
  fail = 1,
  ref_value = NULL,
  strata = NULL,
  per = 1,
  rnd = 4,
  plot = TRUE,
  print.table = TRUE
)

## S3 method for class 'list'
stmh(
  data = NULL,
  time,
  status,
  by,
  fail = 1,
  ref_value = NULL,
  strata = NULL,
  per = 1,
  rnd = 4,
  plot = TRUE,
  print.table = TRUE
)

calc_RR(t, ref_lvl, rnd)

Arguments

data	an optional data frame
time	specify the timer interval when the subject is at risk
status	event of interest to calculate incidence: 1 for event, 0 for censored
by	specify variable to calculate stratified rates
fail	failure event
ref_value	a number or character representing reference value
strata	variable to stratify by Mantel-Haenszel method
per	units to be used in reported rates
rnd	Rounding of numbers
plot	logical value to display plots of rates across a categorical variable
print.table	logical value to display formatted outputs
t	2x2 table input to calaculate odds ratio of non-reference
ref_lvl	logical vector to indicate where reference category is

Details

Rates of event occurrences, known as incidence rates are outcome measures in longitudinal studies. In most longitudinal studies, follow-up times vary due to logistic resasons, different periods of recruitment, delay enrolment into the study, lost-to-follow-up, immigration or emigration and death.

Follow-up time in longitudinal studies

Period of observation (called as follow-up time) starts when individuals join the study and ends when they either have an outcome of interest, are lost-to- follow-up or the follow-up period ends, whichever happens first. This period is called person-year-at-risk. This is denoted by PY in strate function's output and numer of event by D.

Rate Ratios

is calcluated using the following formula:

RR = Rate of Exposured Group / Rate of Non-Exposed Group

Confidence interval of RR

is derived using the following formula:

95\% CI (RR) = Rate x Error Factor

Error Factor (rate) = exp(1.96 / √{1/D1 + 1/D2})

Test of null hypothesis

is calculated as follows:

E1 = ( D * Y1/Y)

where D = (D0 + D1) is the total number of events, and Y= (Y0 + Y1) is the total follow-up time.

The test is based on the difference between the observed number of events in the exposed group and its expected value, D1 – E1 , which we denote by U. The variance of U, this difference, is equal to:

V = D \* (Y1/Y) \* ( 1 – (Y1/Y) )

The square of the standardized difference

U^2 / V

which is referred to the χ2 distribution with 1 degree of freedom (df).

plot, if TRUE, produces a graph of the rate against the numerical code used for categories of by.

Author(s)

Myo Minn Oo (Email: [e-mail] | Website: https://myominnoo.github.io/)

See Also

strate

Examples

## Not run: 
library(lung)
data(lung)
str(lung)

## incidence rates
strate(lung, time, status)

## stratified incidence rates
strate(lung, time, status, sex, per = 100000)
stmh(lung, time, status, sex, per = 100000)

## creating ageband
lung1 <- egen(lung, age, c(55,61,71), new.var = ageband)
strate(lung1, time, status, ageband, per = 100000)
stmh(lung1, time, status, geband, per = 100000)

data(pbc)
str(pbc)
strate(pbc, time, status, trt, per = 100000)
stmh(pbc, time, status, trt, per = 100000)

strate(pbc, time, status, sex, per = 100000)
stmh(pbc, time, status, sex, per = 100000)

## End(Not run)

myominnoo/mStats_beta documentation built on Feb. 29, 2020, 8:17 a.m.