tidying

Signed-off-by: Nathaniel <NathanielF@users.noreply.github.com>

Author: NathanielF

docs/source/notebooks/iv_weak_instruments.ipynb

@@ -5492,7 +5492,7 @@
    "source": [
     "### Credible Inference and the Credibility Revolution.\n",
     "\n",
-    "How far can we push the idea? How vulnerable is our inference to model mis-specification? Can we stress test the parameter estimates by trying strong priors? The IV methodology is so palpably about argument, and credible argument for the proposed mechanism. One benefit of the Bayesian modelling approach to IV is that we can express and stretch the credibility of the mechanism in the model design. We can stress-test our the credibility of inferences by trying to impose contestable beliefs as priors on the model and see the degree to which the inferences are anchored by our prior specifications and how much the data pulls us away from more incredible postulates. \n",
+    "How far can we push the idea? How vulnerable is our inference to model mis-specification? Can we stress test the parameter estimates by trying strong priors? The IV methodology is palpably about argument - credible argument for the proposed mechanism. With this in mind, one benefit of the Bayesian modelling approach to IV is that we can express and stretch the credibility of the mechanism in the model design. We can stress-test the credibility of inferences by trying to impose contestable beliefs as priors on the model and see the degree to which the inferences are anchored by our prior specifications and how much the data pulls us away from more incredible postulates. \n",
     "\n",
     "Here we'll refit our original IV model, but we'll scale the `experience_1` variable to have a mean of 0. This allows place priors of roughly the same scale on all variables. "
    ]
@@ -6042,7 +6042,7 @@
    "source": [
     "### Comparing Model Inferences\n",
     "\n",
-    "Just to make the emerging picture clearer we pull out and plot the credible intervals for the parameter estimates implied by each model. Here we can see the different implications induced by each model configuration. Note how all the IV models pull the credible intervals awat from the simpler OLS type model estimate. "
+    "Just to make the emerging picture clearer we pull out and plot the credible intervals for the parameter estimates implied by each model. Here we can see the different implications induced by each model configuration. Note how all the IV models pull the credible intervals away from the simpler OLS type model estimate. "
    ]
   },
   {
@@ -6282,7 +6282,7 @@
    "cell_type": "markdown",
    "metadata": {},
    "source": [
-    "Our final model shows a somewhat chaotic realisations of the predicted and counterfactual distributions suggesting it is not in fact a great model for our data. We can also formalise this model comparison in their predictive power.  "
+    "Our final model shows a somewhat chaotic realisations of the predicted and counterfactual distributions suggesting it is not in fact a great model for our data. We can also formalise this model comparison in their predictive power."
    ]
   },
   {
@@ -6364,6 +6364,13 @@
     "compare_df[['rank',\t'elpd_loo',\t'p_loo','elpd_diff','weight']]"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "The final analysis should be driven by the plausibility of the implications on the raw `wage` scale. But it's useful to know how and whether a particular model fit compares in predictive power against other reasonable candidate models. "
+   ]
+  },
   {
    "cell_type": "markdown",
    "metadata": {},
@@ -6375,7 +6382,7 @@
    "cell_type": "markdown",
    "metadata": {},
    "source": [
-    "In this demonstration we've seen an example of IV regression justification conducted in a Bayesian setting. Crucially we've tried to convey the centrality of argument and model justification that is required in this mode of causal inference. These requirements are true whether you're fitting the IV model with a frequentist model or a Bayesian model, but the process of justification and model credibility comparison is neatly phrased in the routine of Bayesian model workflow. "
+    "In this demonstration we've seen an example of IV regression justification conducted in a Bayesian setting. Crucially we've tried to convey the centrality of argument and model justification that is required in this mode of causal inference. These requirements are true whether you're fitting the IV model with a frequentist model or a Bayesian model, but the process of justification and model-credibility-comparison is neatly phrased in the routine of Bayesian model workflow. "
    ]
   }
  ],