Analyses of Deviance Summarising Fits of Univariable GLMs

univ_anova() provides a succinct summary of the analyses of deviance for all univariable GLMs based on each possible categorical independent variable in a Bernoulli or binomial data set.

Usage

univ_anova(
  data,
  .dep_var,
  .family = binomial,
  .test = c("none", "Rao", "LRT", "Chisq", "F")
)

Arguments

data: a data frame, or a data frame extension (e.g. a tibble).
.dep_var: <data-masking> quoted name of the binary dependent variable, which should be numeric with values of 0 and 1, or a two-column matrix with the columns giving the numbers of successes and failures e.g., cbind(pn, qn).
.family: a family function; default "binomial". (See family for details of family functions.)
.test: a character string, (partially) matching one of "none", "Chisq", "LRT", "Rao", "F" or "Cp"; default "none".

Value

An object of class c("univ_anova" "announce"), inheriting from "anova" and "data.frame" as for add1(), summarising the differences between the fitted univariable GLMs and the null model.

Details

univ_anova() uses add1() in the stats package to compare univariable GLMs for each possible categorical independent variable (i.e., "factor" columns) in .data with the null model. Other data types e.g. "character" vectors will be ignored and should be converted to "factor" to be included in this analysis.

Examples

(d <- list(
    iv2 = list(g = c("a", "c", "e"), h = c("b", "d", "f")),
    iv3 = list(i = c("a", "b", "c"), j = c("d", "e", "f")),
    iv4 = list(k = c("a", "b"), l = c("c", "d"), m = c("e", "f"))
) |> add_grps(bernoulli_data(levels = 6), iv, .key = _))
#> ___________________________
#> Simulated Bernoulli Data: -
#> 
#> # A tibble: 396 × 5
#>    iv    iv2   iv3   iv4      dv
#>    <fct> <fct> <fct> <fct> <int>
#>  1 a     g     i     k         1
#>  2 a     g     i     k         0
#>  3 a     g     i     k         0
#>  4 a     g     i     k         1
#>  5 a     g     i     k         1
#>  6 a     g     i     k         0
#>  7 a     g     i     k         1
#>  8 a     g     i     k         0
#>  9 a     g     i     k         1
#> 10 a     g     i     k         1
#> # ℹ 386 more rows

d |> univ_anova(dv, .test = "LRT")
#> ___________________________________
#> Univariable Analysis of Deviance: -
#> 
#> Single term additions
#> 
#> Model:
#> dv ~ 1
#>        Df Deviance    AIC    LRT  Pr(>Chi)    
#> <none>      477.19 479.19                     
#> iv      5   423.91 435.91 53.284 2.941e-10 ***
#> iv2     1   470.67 474.67  6.516   0.01069 *  
#> iv3     1   432.95 436.95 44.241 2.903e-11 ***
#> iv4     2   433.06 439.06 44.125 2.620e-10 ***
#> ---
#> Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

d |> binom_contingency(dv, starts_with("iv")) |> print() |>
     univ_anova(cbind(pn, qn), .test = "LRT")
#> _____________________________
#> Binomial Contingency Table: -
#> 
#> # A tibble: 6 × 6
#>   iv    iv2   iv3   iv4      pn    qn
#> * <fct> <fct> <fct> <fct> <int> <int>
#> 1 a     g     i     k        38    28
#> 2 b     h     i     k        26    40
#> 3 c     g     i     l        23    43
#> 4 d     h     j     l        12    54
#> 5 e     g     j     m         8    58
#> 6 f     h     j     m         8    58
#> ___________________________________
#> Univariable Analysis of Deviance: -
#> 
#> Single term additions
#> 
#> Model:
#> cbind(pn, qn) ~ 1
#>        Df Deviance    AIC    LRT  Pr(>Chi)    
#> <none>      53.284 80.822                     
#> iv      5    0.000 37.538 53.284 2.941e-10 ***
#> iv2     1   46.768 76.306  6.516   0.01069 *  
#> iv3     1    9.043 38.580 44.241 2.903e-11 ***
#> iv4     2    9.159 40.696 44.125 2.620e-10 ***
#> ---
#> Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

rm(d)