Busby 47
Roach 14
Bilbray 13
Kaloogian 9
Morrow 5
Uke 3

These numbers are tantalizingly close to Datamar’s, except for Busby, shown here within striking distance of the magic 50%.  The reason is the sample.  Here is the SurveyUSA explanation:

Datamar is relying on past voting history to determine who is a likely voter, including in its survey only voters who voted in both 2004 and 2002. SurveyUSA is relying on the stated intention and the stated passion of poll respondents to determine who is a likely voter. SurveyUSA’s likely voter model detects more enthusiasm among Democrats than among Republicans in CA 50. That is reflected in SurveyUSA’s voter-turnout model, and in the data we present here. Datamar shows 54% of likely voters to be Republican. SurveyUSA shows 45% of likely voters to be Republican. Datamar shows 32% of likely voters to be Democrat. SurveyUSA shows 37% of likely voters to be Democrat. Datamar shows 49% of likely voters to be male. SurveyUSA shows 51% of likely voters to be male. Datamar has 54% of likely voters earning more than $80,000 a year. SurveyUSA has 47% of likely voters earning more than $80,000 a year. The election presents researchers and scholars with an opportunity to learn whether, in a special election, past-voting behavior is more predictive of outcome than stated intent.

See the attachment: