R Graphical Manual

Browse All

Last data update: 2014.03.03

R: Function to calculate effects

effx	R Documentation

Function to calculate effects

Description

The function calculates the effects of an exposure on a response, possibly stratified by a stratifying variable, and/or controlled for one or more confounding variables.

Usage

effx( response, type = "metric",
	         fup = NULL,
	    exposure,
	      strata = NULL,
	     control = NULL,
             weights = NULL,
                 eff = NULL,
	       alpha = 0.05,
	        base = 1,
	      digits = 3,
	        data = NULL )

Arguments

`response`	The `response` variable - must be numeric or logical. If logical, `TRUE` is considered the outcome.
`type`	The type of response`type` - must be one of "metric", "binary", "failure", or "count"
`fup`	The `fup` variable contains the follow-up time for a failure response

. This must be numeric.

`exposure`	The `exposure` variable can be numeric or a factor
`strata`	The `strata` stratifying variable - must be a factor
`control`	The `control` variable(s) (confounders) - these are passed as a list if there are more than one.
`weights`	Frequency weights for binary response only
`eff`	How should effects be measured. If `response` is binomial, the default is "OR" (odds-ratio) with "RR" (relative risk) as an option. If `response` is failure, the default is "RR" (rate-ratio) with "RD" (rate difference) as an option.
`base`	Baseline for the effects of a categorical exposure, either a number or a name of the level. Defaults to 1
`digits`	Number of significant digits for the effects, default 3
`alpha`	1 - confidence level
`data`	`data` refers to the data used to evaluate the function

Details

The function is a wrapper for glm. Effects are calculated as differences in means for a metric response, odds ratios/relatiev risks for a binary response, and rate ratios/rate differences for a failure or count response.

The k-1 effects for a categorical exposure with k levels are relative to a baseline which, by default, is the first level. The effect of a metric (quantitative) exposure is calculated per unit of exposure.

The exposure variable can be numeric or a factor, but if it is an ordered factor the order will be ignored.

Value

`comp1`	Effects of exposure
`comp2`	Tests of significance

Author(s)

Michael Hills

References

www.mhills.pwp.blueyonder.co.uk

Examples

library(Epi)
data(births)
births$hyp <- factor(births$hyp,labels=c("normal","hyper"))
births$sex <- factor(births$sex,labels=c("M","F"))

# bweight is the birth weight of the baby in gms, and is a metric
# response (the default)

# effect of hypertension on birth weight
effx(bweight,exposure=hyp,data=births)
# effect of hypertension on birth weight stratified by sex
effx(bweight,exposure=hyp,strata=sex,data=births)
# effect of hypertension on birth weight controlled for sex
effx(bweight,exposure=hyp,control=sex,data=births)
# effect of gestation time on birth weight
effx(bweight,exposure=gestwks,data=births)
# effect of gestation time on birth weight stratified by sex
effx(bweight,exposure=gestwks,strata=sex,data=births)
# effect of gestation time on birth weight controlled for sex
effx(bweight,exposure=gestwks,control=sex,data=births)

# lowbw is a binary response coded 1 for low birth weight and 0 otherwise
# effect of hypertension on low birth weight
effx(lowbw,type="binary",exposure=hyp,data=births)
effx(lowbw,type="binary",exposure=hyp,eff="RR",data=births)

Results


R version 3.3.1 (2016-06-21) -- "Bug in Your Hair"
Copyright (C) 2016 The R Foundation for Statistical Computing
Platform: x86_64-pc-linux-gnu (64-bit)

R is free software and comes with ABSOLUTELY NO WARRANTY.
You are welcome to redistribute it under certain conditions.
Type 'license()' or 'licence()' for distribution details.

R is a collaborative project with many contributors.
Type 'contributors()' for more information and
'citation()' on how to cite R or R packages in publications.

Type 'demo()' for some demos, 'help()' for on-line help, or
'help.start()' for an HTML browser interface to help.
Type 'q()' to quit R.

> library(Epi)

Attaching package: 'Epi'

The following object is masked from 'package:base':

    merge.data.frame

> png(filename="/home/ddbj/snapshot/RGM3/R_CC/result/Epi/effx.Rd_%03d_medium.png", width=480, height=480)
> ### Name: effx
> ### Title: Function to calculate effects
> ### Aliases: effx
> ### Keywords: models regression
> 
> ### ** Examples
> 
> library(Epi)
> data(births)
> births$hyp <- factor(births$hyp,labels=c("normal","hyper"))
> births$sex <- factor(births$sex,labels=c("M","F"))
> 
> # bweight is the birth weight of the baby in gms, and is a metric
> # response (the default)
> 
> # effect of hypertension on birth weight
> effx(bweight,exposure=hyp,data=births)
--------------------------------------------------------------------------- 
response      :  bweight 
type          :  metric 
exposure      :  hyp 

hyp is a factor with levels: normal / hyper 
baseline is  normal 
effects are measured as differences in means 
--------------------------------------------------------------------------- 

effect of hyp on bweight 
number of observations  500 

Effect   2.5%  97.5% 
  -431   -585   -276 

Test for no effects of exposure on 1 df: p-value= 4.9e-08 
> # effect of hypertension on birth weight stratified by sex
> effx(bweight,exposure=hyp,strata=sex,data=births)
--------------------------------------------------------------------------- 
response      :  bweight 
type          :  metric 
exposure      :  hyp 
stratified by :  sex 

hyp is a factor with levels: normal / hyper 
baseline is  normal 
sex is a factor with levels: M/F 
effects are measured as differences in means 
--------------------------------------------------------------------------- 

effect of hyp on bweight 
stratified by sex 

number of observations  500 

                               Effect 2.5% 97.5%
strata M level hyper vs normal   -496 -696  -297
strata F level hyper vs normal   -380 -617  -142

Test for effect modification on 1 df: p-value= 0.462 
> # effect of hypertension on birth weight controlled for sex
> effx(bweight,exposure=hyp,control=sex,data=births)
--------------------------------------------------------------------------- 
response      :  bweight 
type          :  metric 
exposure      :  hyp 
control vars  :  sex 

hyp is a factor with levels: normal / hyper 
baseline is  normal 
effects are measured as differences in means 
--------------------------------------------------------------------------- 

effect of hyp on bweight 
controlled for sex 

number of observations  500 

Effect   2.5%  97.5% 
  -448   -601   -295 

Test for no effects of exposure on 1 df: p-value= 9.07e-09 
> # effect of gestation time on birth weight
> effx(bweight,exposure=gestwks,data=births)
--------------------------------------------------------------------------- 
response      :  bweight 
type          :  metric 
exposure      :  gestwks 

gestwks is numeric 
effects are measured as differences in means 
--------------------------------------------------------------------------- 

effect of an increase of 1 unit in gestwks on bweight 
number of observations  490 

Effect   2.5%  97.5% 
   197    180    214 

Test for no effects of exposure on 1 df: p-value= <2e-16 
> # effect of gestation time on birth weight stratified by sex
> effx(bweight,exposure=gestwks,strata=sex,data=births)
--------------------------------------------------------------------------- 
response      :  bweight 
type          :  metric 
exposure      :  gestwks 
stratified by :  sex 

gestwks is numeric 
sex is a factor with levels: M/F 
effects are measured as differences in means 
--------------------------------------------------------------------------- 

effect of an increase of 1 unit in gestwks on bweight 
stratified by sex 

number of observations  490 

         Effect 2.5% 97.5%
strata M    203  177   229
strata F    191  169   213

Test for effect modification on 1 df: p-value= 0.487 
> # effect of gestation time on birth weight controlled for sex
> effx(bweight,exposure=gestwks,control=sex,data=births)
--------------------------------------------------------------------------- 
response      :  bweight 
type          :  metric 
exposure      :  gestwks 
control vars  :  sex 

gestwks is numeric 
effects are measured as differences in means 
--------------------------------------------------------------------------- 

effect of an increase of 1 unit in gestwks on bweight 
controlled for sex 

number of observations  490 

Effect   2.5%  97.5% 
   196    179    213 

Test for no effects of exposure on 1 df: p-value= <2e-16 
> 
> # lowbw is a binary response coded 1 for low birth weight and 0 otherwise
> # effect of hypertension on low birth weight
> effx(lowbw,type="binary",exposure=hyp,data=births)
--------------------------------------------------------------------------- 
response      :  lowbw 
type          :  binary 
exposure      :  hyp 

hyp is a factor with levels: normal / hyper 
baseline is  normal 
effects are measured as odds ratios 
--------------------------------------------------------------------------- 

effect of hyp on lowbw 
number of observations  500 

Effect   2.5%  97.5% 
  3.73   2.03   6.87 

Test for no effects of exposure on 1 df: p-value= 6.06e-05 
> effx(lowbw,type="binary",exposure=hyp,eff="RR",data=births)
--------------------------------------------------------------------------- 
response      :  lowbw 
type          :  binary 
exposure      :  hyp 

hyp is a factor with levels: normal / hyper 
baseline is  normal 
effects are measured as relative risk 
--------------------------------------------------------------------------- 

effect of hyp on lowbw 
number of observations  500 

Effect   2.5%  97.5% 
  2.97   1.85   4.78 

Test for no effects of exposure on 1 df: p-value= 6.06e-05 
> 
> 
> 
> 
> 
> dev.off()
null device 
          1 
>