# pseudoyl: Pseudo-observations for the expected number of years lost In pseudo: Computes Pseudo-Observations for Modeling

## Description

Computes pseudo-observations for modeling using the number of years lost.

## Usage

 `1` ```pseudoyl(time,event, tmax) ```

## Arguments

 `time` the follow up time. `event` the cause indicator, use 0 as censoring code and integers to name the other causes. `tmax` the maximum cut-off point time = the upper limit of the integral of the cumulative incidence function. If missing or larger than the maximum follow up time, it is replaced by the maximum follow up time.

## Details

The function calculates the pseudo-observations for the expected number of years lost for each individual. The pseudo-observations can be used for fitting a regression model with a generalized estimating equation. No missing values in either `time` or `event` vector are allowed.

## Value

A list containing the following objects:

 `cause` The ordered codes for different causes. `pseudo` A list of vectors- a vector for each of the causes, ordered by codes. Each value of a vector belongs to one individual (ordered as in the original data set).

## References

Andersen P.K.: "A note on the decomposition of number of life years lost according to causes of death." Research report, Department of Biostatistics, University of Copenhagen, 2012 (2)

`pseudoci`, `pseudomean`, `pseudosurv`

## Examples

 ``` 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23``` ```library(KMsurv) data(bmt) bmt\$icr <- bmt\$d1 + bmt\$d3 #compute the pseudo-observations: pseudo = pseudoyl(time=bmt\$t2, event=bmt\$icr,tmax=2000) #arrange the data - use pseudo observations for cause 2 a <- cbind(bmt,pseudo = pseudo\$pseudo[[2]],id=1:nrow(bmt)) #fit a regression model for cause 2 library(geepack) summary(fit <- geese(pseudo ~ z1 + as.factor(z8) + as.factor(group), data = a, id=id, jack = TRUE, family=gaussian, corstr="independence", scale.fix=FALSE)) #rearrange the output round(cbind(mean = fit\$beta,SD = sqrt(diag(fit\$vbeta.ajs)), Z = fit\$beta/sqrt(diag(fit\$vbeta.ajs)), PVal = 2-2*pnorm(abs(fit\$beta/sqrt(diag(fit\$vbeta.ajs))))),4) ```

### Example output

```Loading required package: KMsurv

Call:
geese(formula = pseudo ~ z1 + as.factor(z8) + as.factor(group),
id = id, data = a, family = gaussian, scale.fix = FALSE,
corstr = "independence", jack = TRUE)

Mean Model:
Variance to Mean Relation: gaussian

Coefficients:
estimate     san.se     ajs.se       wald            p
(Intercept)        879.34078 219.365201 223.196421 16.0686341 6.108751e-05
z1                  10.10276   6.904353   7.083291  2.1410852 1.434004e-01
as.factor(z8)1     496.34909 171.529255 174.862683  8.3733271 3.807679e-03
as.factor(group)2 -673.50096 184.882614 186.715577 13.2704084 2.696285e-04
as.factor(group)3 -119.96162 218.358110 222.249313  0.3018186 5.827446e-01

Scale Model:

Estimated Scale Parameters:
estimate   san.se   ajs.se     wald p
(Intercept)   636758 49853.98 52685.27 163.1357 0

Correlation Model:
Correlation Structure:     independence

Returned Error Value:    0
Number of clusters:   137   Maximum cluster size: 1

mean       SD       Z   PVal
(Intercept)        879.3408 223.1964  3.9398 0.0001
z1                  10.1028   7.0833  1.4263 0.1538
as.factor(z8)1     496.3491 174.8627  2.8385 0.0045
as.factor(group)2 -673.5010 186.7156 -3.6071 0.0003
as.factor(group)3 -119.9616 222.2493 -0.5398 0.5894
```

pseudo documentation built on May 1, 2019, 6:35 p.m.