Dependent (Response) Variable Extracted

A generic function that extracts the response/dependent variable from objects.

Usage

depvar(object, ...)

Arguments

object

An object that has some response/dependent variable.

...

Other arguments fed into the specific methods function of the model. The following mainly applies to "vglm" class objects. In particular, sometimes type = c("lm", "lm2") is available, in which case the first one is chosen if the user does not input a value. The latter value corresponds to argument form2, and sometimes a response for that is optional.

Another argument is muxypw (whose default is FALSE) which is whether object@y needs to be multiplied by the prior weight. This feature is useful for family functions that store sample proportions in object@y and the prior weights in object@prior.weights because sometimes the count response is needed, e.g., for vglmnet(). So if on object@family@infos() muxypw == FALSE and if Mux4vglmnet = TRUE then set depvar(object, muxypw = TRUE) to get the raw counts. Also, because roundmux = FALSE is the default, it is a good idea to set roundmux = TRUE to make sure the result is integer-valued, i.e., depvar(object, muxypw = TRUE, roundmux = TRUE). Note that recycling is used.

Details

By default this function is preferred to calling object@y, say.

Value

The response/dependent variable, usually as a matrix or vector.

Author

Thomas W. Yee

Examples

pneumo <- transform(pneumo, let = log(exposure.time))
(fit <- vglm(cbind(normal, mild, severe) ~ let, propodds, pneumo))
#> 
#> Call:
#> vglm(formula = cbind(normal, mild, severe) ~ let, family = propodds, 
#>     data = pneumo)
#> 
#> 
#> Coefficients:
#> (Intercept):1 (Intercept):2           let 
#>     -9.676093    -10.581725      2.596807 
#> 
#> Degrees of Freedom: 16 Total; 13 Residual
#> Residual deviance: 5.026826 
#> Log-likelihood: -25.09026 
fit@y        # Sample proportions (not recommended)
#>      normal       mild     severe
#> 1 1.0000000 0.00000000 0.00000000
#> 2 0.9444444 0.03703704 0.01851852
#> 3 0.7906977 0.13953488 0.06976744
#> 4 0.7291667 0.10416667 0.16666667
#> 5 0.6274510 0.19607843 0.17647059
#> 6 0.6052632 0.18421053 0.21052632
#> 7 0.4285714 0.21428571 0.35714286
#> 8 0.3636364 0.18181818 0.45454545
depvar(fit)  # Better than using fit@y
#>      normal       mild     severe
#> 1 1.0000000 0.00000000 0.00000000
#> 2 0.9444444 0.03703704 0.01851852
#> 3 0.7906977 0.13953488 0.06976744
#> 4 0.7291667 0.10416667 0.16666667
#> 5 0.6274510 0.19607843 0.17647059
#> 6 0.6052632 0.18421053 0.21052632
#> 7 0.4285714 0.21428571 0.35714286
#> 8 0.3636364 0.18181818 0.45454545
weights(fit, type = "prior")  # Number of observations
#>   [,1]
#> 1   98
#> 2   54
#> 3   43
#> 4   48
#> 5   51
#> 6   38
#> 7   28
#> 8   11
# This is new:
fit@family@infos()$muxypw  # Proportion or count?
#> [1] FALSE
fit@family@infos()$Mux4vglmnet  # vglmnet needs counts?
#> [1] TRUE
fit@family@infos()$roundmux  # Product is integer-valued?
#> [1] TRUE
depvar(fit, muxypw = TRUE, roundmux = TRUE)  # Counts
#>   normal mild severe
#> 1     98    0      0
#> 2     51    2      1
#> 3     34    6      3
#> 4     35    5      8
#> 5     32   10      9
#> 6     23    7      8
#> 7     12    6     10
#> 8      4    2      5