Terrible Statistics Advice

Fall in love with your models.

The Central Limit Theorem probably applies to data drawn from a Cauchy distribution, whatever the naysayers may argue.

Use Microsoft Excel to create all your charts.

Yeah, that time series is probably stationary already, don't worry about it.

Make yourself indispensable by making all your statistical analysis impossible to replicate.

Save time and effort by training your model on the test data.

Never normalize values. Units are important.

Don’t worry so much about measurement error or missing data.

Condition on a collider variable.

Create a chart with a secondary vertical axis.

Collect the data first, then worry about experimental design.

Add more regressors to your model to increase R^2.

Confounders are not real and cannot hurt you.

Focus all your efforts on calculating point estimates.

Illustrate a discrete time series using a chart with smooth lines.

Remove all outliers before attempting any analysis.