The evidence provided by a single trial is less reliable than its statistical analysis suggests