# Chapter 9 | Exercise 4

Which register has the greatest proportion of metaphoric uses of *on*? The smallest? The frequencies in the VU Amsterdam Metaphor Corpus are shown in Table 9.1.

Academic | Conversations | Fiction | News | |
---|---|---|---|---|

Metaphoric | 255 | 163 | 112 | 205 |

Non-metaphoric | 37 | 96 | 121 | 77 |

Test if there is a significant association between the register and (non)-metaphoricity. Visualize the data in a mosaic and association plot.

First, create a table in R with all frequencies:

```
> met <- c(255, 163, 112, 205)
> nonmet <- c(37, 96, 121, 77)
> on <- rbind(met, nonmet)
> colnames(on) <- c('acad', 'conv', 'fic', 'news')
> on
acad conv fic news
met 255 163 112 205
nonmet 37 96 121 77
```

To compare the proportions of metaphoric and non-metaphoric uses in four registers, you can use the following code:

```
> prop.table(on, 2)
acad conv fic news
met 0.8732877 0.6293436 0.4806867 0.7269504
nonmet 0.1267123 0.3706564 0.5193133 0.2730496
```

The academic register has the greatest proportion of metaphoric uses (0.87). Fiction has the lowest proportion (0.48).

Now we should test if the variables are associated:

```
> chisq.test(on)
Pearson's Chi-squared test
data: on
X-squared = 99.7487, df = 3, p-value < 2.2e-16
```

The *p*-value is almost zero, so the null hypothesis of no association can be safely discarded. To check the contribution of each cell to the result, check the residuals:

```
> chisq.test(on)$residuals
acad conv fic news
met 3.78232 -1.165787 -3.838463 0.7575156
nonmet -5.63622 1.737196 5.719882 -1.1288110
```

The positive and negative values of the residuals indicate the same differences that we have observed. To visualize the residuals, use the `mosaic()`

and `assoc()`

functions in the `vcd`

package:

```
> library(vcd)
> mosaic(on, shade = T)
> assoc(on, shade = T)
```