# crrstep-package: Stepwise regression procedure for the Fine & Gray regression... In crrstep: Stepwise Covariate Selection for the Fine & Gray Competing Risks Regression Model

## Description

Performs forward and backward stepwise regression for the Fine & Gray regression model in competing risks. Procedure uses AIC, BIC and BICcr as selection criteria. BICcr has a penalty of k = log(n*), where n* is the number of Type I events.

## Details

 Package: crrstep Type: Package Version: 2014-07.16 Date: 2014-07.16 License: GPL (version 2) LazyLoad: yes

The package contains a singe function `crrstep`, which implements backward and forward stepwise regression for the Fine & Gray regression model. The Fine & Gray model (Fine & Gray, 1999) estimates the hazard that corresponds to the cumulative incidence function of a certain event type. Selection criteria that are can be used are: AIC, BIC and BICcr. BICcr is a selection criteria based on the work by Volinksy and Raftery in which the penalty is k = log(n*), where n* is the total number of Type I events.

## References

Fine, J. P. and Gray, R. J. (1999). A proportional hazards model for the subdistribution of a competing risk. Journal of the American Statistical Association.

Volinsky, C. T. and Raftery, A. E. (2000). Bayesian information criterion for censored survival models. Biometrics.

Kuk, D. and Varadhan, R. (2013). Model selection in competing risks regression. Statistics in Medicine.

## Examples

 ``` 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17``` ```set.seed(123) n <- 500 ftime <- rexp(n) fstatus <- sample(0:2,n,replace=TRUE) cov1 <- matrix(runif(5*n),nrow=n) x61 <- as.factor(sample(3, size=n, rep=TRUE)) x71 <- as.factor(sample(5, size=n, rep = TRUE)) cov1 <- cbind(cov1, x61, x71) dimnames(cov1)[[2]] <- c('x1','x2','x3','x4','x5', 'x6', 'x7') formula1 <- ftime ~ 1 + x1 + x2 + x3 + x4 + x5 + as.factor(x6) + as.factor(x7) crrstep(formula1, , fstatus, data = as.data.frame(cov1), direction = "backward", criterion = "BIC") crrstep(formula1, , fstatus, data = as.data.frame(cov1), direction = "backward", criterion = "AIC") ans2 <- crrstep(formula1, , fstatus, data = as.data.frame(cov1), direction = "forward", failcode=2, criterion = "AIC") ans2 ```

### Example output

```Loading required package: cmprsk
crrstep(formula = formula1, etype = fstatus, data = as.data.frame(cov1),
direction = "backward", criterion = "BIC")
BIC
-as.factor(x7) 2019.46
-as.factor(x6) 2031.49
-x2            2037.61
-x1            2037.66
-x3            2037.75
-x4            2038.72
-x5            2042.07
<none>         2043.62
[1] "as.factor(x7)"

Step: BIC= 2019.46
crr(ftime = ftime, fstatus = fstatus, cov1 = cov3.2, variance = TRUE)

BIC
-as.factor(x6) 2007.41
-x2            2013.46
-x1            2013.51
-x3            2013.65
-x4            2014.50
-x5            2017.91
<none>         2019.46
[1] "as.factor(x6)"

Step: BIC= 2007.41
crr(ftime = ftime, fstatus = fstatus, cov1 = cov3.2, variance = TRUE)

BIC
-x2    2001.38
-x1    2001.46
-x3    2001.62
-x4    2002.50
-x5    2005.73
<none> 2007.41
[1] "x2"

Step: BIC= 2001.38
crr(ftime = ftime, fstatus = fstatus, cov1 = cov3.2, variance = TRUE)

BIC
-x1    1995.42
-x3    1995.67
-x4    1996.44
-x5    1999.63
<none> 2001.38
[1] "x1"

Step: BIC= 1995.42
crr(ftime = ftime, fstatus = fstatus, cov1 = cov3.2, variance = TRUE)

BIC
-x3    1989.76
-x4    1990.52
-x5    1993.50
<none> 1995.42
[1] "x3"

Step: BIC= 1989.76
crr(ftime = ftime, fstatus = fstatus, cov1 = cov3.2, variance = TRUE)

BIC
-x4    1984.91
-x5    1988.12
<none> 1989.76
[1] "x4"

Step: BIC= 1984.91
crr(ftime = ftime, fstatus = fstatus, cov1 = cov3.2, variance = TRUE)

BIC
-x5    1983.45
<none> 1984.91
[1] "x5"
[1] "The best model is the NULL model and the likelihood is:-991.73"

Step: BIC= 1978.7
crr(ftime = ftime, fstatus = fstatus, cov1 = cov3.2, variance = TRUE)

\$coefficients
NULL

\$log.likelihood
[1] -989.35

\$coefficients
NULL

\$log.likelihood
[1] -989.35

crrstep(formula = formula1, etype = fstatus, data = as.data.frame(cov1),
direction = "backward", criterion = "AIC")
AIC
-as.factor(x7) 1989.96
-as.factor(x6) 1993.56
-x2            1995.46
-x1            1995.51
-x3            1995.61
-x4            1996.58
<none>         1997.26
-x5            1999.92
[1] "as.factor(x7)"

Step: AIC= 1989.96
crr(ftime = ftime, fstatus = fstatus, cov1 = cov3.2, variance = TRUE)

AIC
-as.factor(x6) 1986.34
-x2            1988.17
-x1            1988.22
-x3            1988.36
-x4            1989.21
<none>         1989.96
-x5            1992.62
[1] "as.factor(x6)"

Step: AIC= 1986.34
crr(ftime = ftime, fstatus = fstatus, cov1 = cov3.2, variance = TRUE)

AIC
-x2    1984.52
-x1    1984.60
-x3    1984.76
-x4    1985.64
<none> 1986.34
-x5    1988.87
[1] "x2"

Step: AIC= 1984.52
crr(ftime = ftime, fstatus = fstatus, cov1 = cov3.2, variance = TRUE)

AIC
-x1    1982.78
-x3    1983.03
-x4    1983.79
<none> 1984.52
-x5    1986.99
[1] "x1"

Step: AIC= 1982.78
crr(ftime = ftime, fstatus = fstatus, cov1 = cov3.2, variance = TRUE)

AIC
-x3    1981.33
-x4    1982.09
<none> 1982.78
-x5    1985.08
[1] "x3"

Step: AIC= 1981.33
crr(ftime = ftime, fstatus = fstatus, cov1 = cov3.2, variance = TRUE)

AIC
-x4    1980.70
<none> 1981.33
-x5    1983.90
[1] "x4"

Step: AIC= 1980.7
crr(ftime = ftime, fstatus = fstatus, cov1 = cov3.2, variance = TRUE)

AIC
<none> 1980.70
-x5    1983.45

\$coefficients
estimates std.error t-stat
x5     0.565     0.253   2.23

\$log.likelihood
[1] -989.35

\$coefficients
estimates std.error t-stat
x5     0.565     0.253   2.23

\$log.likelihood
[1] -989.35

crrstep(formula = formula1, etype = fstatus, failcode = 2, data = as.data.frame(cov1),
direction = "forward", criterion = "AIC")
NULL
AIC
<none>         1794.03
+as.factor(x7) 1794.52
+x1            1795.73
+x4            1795.79
+x2            1795.96
+x3            1796.01
+x5            1796.02
+as.factor(x6) 1796.29
[1] "The best model is the NULL model and the likelihood is:-897.01"
\$coefficients
NULL

\$log.likelihood
[1] -897.01
```

crrstep documentation built on May 2, 2019, 6:37 a.m.