Suppose that an investigator is conducting a study of political attitudes among undergraduate students at a liberal arts college in central Virginia.

She hypothesizes that students majoring in thehumanitieswill havethe most conservative political attitudes, students majoring in thesocial scienceswill have theleast conservative attitudes, and students majoring inscience, technology, engineering, and mathematicsareas will haveintermediate attitudes.

She conducts a survey of students and, for each, obtains a measure of adherence to conservative political attitudes (CONATT in her dataset, which takes higher values for higher levels of political conservatism), and also asks each student about her or his major.

On the basis of students' majors, she creates three dummy variables:

• DUMHUM, which takes the value 1 for students enrolled in humanities majors
• DUMSOC, which takes the value 1 for students enrolled in social science majors
• DUMSTEM, which takes the value 1 for students enrolled in science, technology, engineering, and math majors

All majors can be classified into one of these three groups. She then tries to run the following linear regression model in SPSS:

E[CONATTIi] =0+1DUMHUMi+2DUMSOCi+3DUMSTEMi

What is wrong with this setup?

She decides to designate science, technology, engineering, and math majors the reference group, and therefore uses SPSS to estimate the following linear regression model:

E[CONATTIi] =0+1DUMHUMi+2DUMSOCi

How should she interpret the estimate of0she obtains from SPSS?