As part of the Perceptions of Women in Academic Scientist (PWAS) study, we selected a random sample of 3,455 scientists from the more than 14,000 graduate students, postdoctoral fellows, and tenure-track/tenured faculty members in the top 20 Ph.D. programs in all subfields of astronomy, physics and biology. Programs were ranked by the National Research Council (1995) and correlated with the rankings of U.S. News
& World Report
(2008). The PWAS survey ran from November 2008 through February 2009, using Web and phone surveys. The sample was stratified by rank in the career track and, where possible, we selected a disproportionately high sample of women within each rank. A personalized contact letter was initially sent to each of the potential respondents. It included a pre-incentive (cash that could be kept regardless of whether the respondent agreed to participate). Respondents were then contacted by a survey firm, Shulman, Ronca, and Buchavales, Inc. (SRBI), who emailed each potential respondent between eight and ten times. Each potential respondent received a unique ID with which to log into a Website and complete the survey. After the reminder emails, SRBI followed up with non-responders through a phone-calling effort, phoning them up to 20 times. Only 7.5 percent of the respondents completed the survey on the phone; 92 percent completed the Web-based survey. Overall, this combination of methods resulted in a high response rate of 72 percent or 2,503 respondents. This is a very high response rate for a survey of academics. For example, even the highly successful Carnegie Commission study of faculty resulted in only a 59.8 percent rate 
. The final sample included 684 graduate students, 504 postdoctoral fellows, 446 assistant professors, 326 associate professors, and 543 full professors. This is an effective response of 75.3 percent among graduate students, 70.7 percent among postdoctoral fellows and 71.7 percent among faculty. An average of 81 persons from each university responded, roughly 25 from each department. A total of 1,300 biologists and 1,203 physicists responded to the study. (A complete guide to the survey questions is available upon request.)
For this paper, questions about significant influences on the pursuit of a science career, career experiences and impediments, and the work-family balance were analyzed from the PWAS survey. Some questions on the PWAS were replicated from existing studies of women in science, such as NSF ADVANCE surveys conducted by PIs examining gender differences in science at particular universities. We used two-tailed tests of statistical significance and logistic regression to analyze the data. T-tests are a premature measure of interpretation because they do not take into account any additional differences between men and women. Logistic regression allows estimates of how one variable predicts another variable, while adjusting for multiple other predictor variables. Thus, logistic regression permits comparison of men and women on key variables of interest while statistically equalizing their different demographic and human capital characteristics. Multivariate logistic regression analyses were weighted by rank, sex, and discipline to reflect chance of selection into the sample. Logistic regression coefficients are reported as odds ratios that center on the number 1. Values less than 1 have a negative effect on the dependent variables while values larger than 1 have a positive effect on the dependent variable. All analyses were conducted with the statistical program STATA version 9 from StataCorp, College Station, TX, USA.
The Perceptions of Women in Academic Science received an expedited human subjects' approval 08/07/08 from Rice University's Institutional Review Board, which has been renewed every year of the three years of data collection (protocol number 09-008E). The PI requested and received permission that informed consent be waived for the survey portion of the study (data from which is reported in this paper). The letter that was sent requesting participation in the survey included an information sheet about the study as well as contact information for the PI and the director of the Rice University IRB should any complaints arise.