docs(count-regression): use estimator macros and fix em-dash spacing

Addresses the three findings carried unaddressed across review rounds:

- `\hat{\vec{\beta}}` at two sites hand-expanded a macro that already
  exists. Now `\evb`, the `\e...` estimator-indirection form CLAUDE.md
  prefers for new content. Verified identical output: both expand to
  `\hat{\tilde{\beta}}` in the render.
- The em dash before a soft line break rendered closed-left/open-right
  once Pandoc collapsed the newline to a space. Moving the dashes to the
  start of the continuation line restores symmetric spacing, confirmed in
  the rendered HTML: "overdispersion -- without ... estimates -- is".

Swept the same class across the PR's own new content, which turned up a
third instance the findings did not name: `\hat{\theta}` -> `\eth`, and
`\hat{\P}` -> `\est{\P}` in prose this PR rewrote.

Four further raw `\hat` sites in `_sec_poisson_dx.qmd` are pre-existing
on main and only reflowed here, so converting one of them would have been
an inconsistent partial sweep of untouched content. Filed as #1144
instead.

Local gate: `quarto render chapters/count-regression.qmd --to html`
succeeded with zero KaTeX errors; `lintr::lint()` clean on both files;
spellcheck reports only pre-existing terms, none introduced here.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

Co-authored-by: d-morrison <2474437+d-morrison@users.noreply.github.com>

Author: d-morrison

_subfiles/count-regression/_sec-overdispersion.qmd

@@ -18,14 +18,14 @@ substantially larger than the mean
 A random variable $Y$ is **overdispersed**
 relative to a parametric model $\P(Y=y \mid \vX)$ if
 its conditional empirical variance in a dataset exceeds
-the theoretical variance imposed by the fitted model $\hat{\P}(Y=y \mid \vX)$.
+the theoretical variance imposed by the fitted model $\est{\P}(Y=y \mid \vX)$.
 
 ::::
 
 In Poisson regression, unmodeled heterogeneity, clustering,
 or omitted predictors cause overdispersion.
 When overdispersion is present but ignored,
-the point estimates $\hat{\vec{\beta}}$ remain unbiased,
+the point estimates $\evb$ remain unbiased,
 but the standard errors produced by standard maximum likelihood estimation
 are severely underestimated.
 This underestimation leads to overly narrow confidence intervals

chapters/count-regression.qmd

@@ -103,18 +103,18 @@ and variance expansion in count data.
 
 ## Quasipoisson regression
 
-Another flexible approach to address overdispersion—
-without changing the underlying Poisson point estimates—
+Another flexible approach to address overdispersion
+— without changing the underlying Poisson point estimates —
 is quasi-likelihood estimation via the quasipoisson model family.
 Rather than specifying a complete likelihood function
 and fitting by maximum likelihood,
 quasipoisson models specify only the mean-variance relationship
 $\Var{Y \mid \vX} = \theta \mu(\vx, t)$,
 where $\theta$ is a dispersion parameter estimated from Pearson residuals.
 
-While point estimates for regression coefficients $\hat{\vec{\beta}}$
+While point estimates for regression coefficients $\evb$
 remain identical to standard Poisson regression,
-their estimated standard errors are scaled by $\sqrt{\hat{\theta}}$.
+their estimated standard errors are scaled by $\sqrt{\eth}$.
 This method-of-moments approach provides robust standard error estimates
 and valid $p$-values when overdispersion is multiplicative,
 though it does not specify a full parametric distribution