As a frequent recipient of baseless criticisms of theoretical models, the most frustrating (yet common) one that I encounter goes: "These results don't fit my own perception/intuition/observations/these data so it can't be true. 1. Models are models; they aim to be useful/insightful, not true. 1/n
2. You need to understand what the model does before criticizing the results it generates. For that, you need to evaluate the model assumptions, not any specific results under given parameter settings. First make sure you get why the model produces certain outcomes. 2/n
Then critically evaluate the viability of the model assumptions (under specific conditions) and compare them against your own assumptions.
3. Never assume your perceptions/intuitions/observations/data are generated under idealized conditions and perfect assumptions of statistical theory. 3/n
Be aware of potential consequences model misspecification, measurement error, confounded/imprecise designs, etc. affecting your inferences. Theoretical models are often idealizations. No direct comparison to real observations are warranted. Use them as intended.
4. Be kind; assume good faith. 4/4
I'll have to read the piece but that highlighted sentence and especially the use of the word "legitimate" rings very wrong to me. It's like saying "confirmation bias is legitimate" which would be weird. Yes it may be a rational move in some situations but not necessarily what we're generally after.
