We must consider for all of the variables the possibility of typeI errors and, consequently,
the existence of spurious associations. We did not consider the existence of the observer because all interviews were performed by a singleauth or without consideration of a particular farm’s serological results. On the other hand, although the questionnaire was previously validated, not all of the answers given bythe farmers could be confirmed by the observations of the author, who sometimes questioned the reliability of the answers. This fact could have resulted in bias.