they said fvap does not conduct a nonresponse bias analysis that omb says is a necessary step in determining whether survey findings are biased. not conducting such an analysis limits data reliability. now i don't want to get into a statistical debate here, because statistics is boring. but the point is anybody who does surveys knows that there is bias in who actually sends them back and how accurate the information is when it comes back. the other problem with the samples were that it didn't accurately reflect the overall military population. for example, 25% of the surveys that came back were completed by senior officers and their spouses, even though there are only about 12% of the overall military population. >> but in making that assertion, as i understand, look, surveys are never perfect. >> i know that. but the point is -- >> compare -- wasn't it weighted? didn't they weight that. >> the gao says they didn't do it, okay. i would argue that we don't have significant evidence of improvement. in fact, the other data from the ec doesn't. i tell you, one of the biggest problems we had in t