Table 2.
Number of parameters, log-likelihood, and Akaike Information Criterion for the 5 models that we tested for the corpus of original tweets created by the @CDC_AMD Twitter profile.
| Model | A | B | C | D | E |
| Model choice | Poisson | Negative binomial | Negative binomial | Hurdlea | Hurdlea |
| Predictorsb | Media+Content | Media+Content | Media only | Media+Content | Media only |
| Number of parameters | 8 | 9 | 4 | 16 | 6 |
| Log-likelihood (dfc) | −1210.125 (df=7) | −1080.27 (df=8) | −1084.925 (df=3) | −1075.625 (df=15) | −1084.567 (df=5) |
| Akaike Information Criterion | 2434.25 | 2176.5 | 2175.9 | 2181.25 | 2179.133 |
aHurdle model (count=negative binomial; zero hurdle=logistic).
bMedia: attachment of a photo or a video (or a link to a photo or a video).
cdf: degrees of freedom.