Combining Categorical Variables

[Robert Michael]

Hello,

When I designed my survey for my research project I had a question that was repeated. Since data was collected on both questions I wanted to combine the two variables (Q3_3 and Q3_4) into one variable. Does anyone have any ideas on how to do this? I have provided the frequency tables below. Thanks in advance for the help.

. tab Q3_3

I believe it is |
important to... - |
provide copies of |
my overhead |
and/or PowerPoin | Freq. Percent Cum.
------------------+-----------------------------------
Strongly disagree | 9 5.92 5.92
Disagree | 8 5.26 11.18
Somewhat disagree | 4 2.63 13.82
Somewhat agree | 16 10.53 24.34
Agree | 29 19.08 43.42
Strongly agree | 86 56.58 100.00
------------------+-----------------------------------
Total | 152 100.00

. tab Q3_4

I believe it is |
important to... - |
provide copies of |
my overhead |
and/or PowerPoin | Freq. Percent Cum.
------------------+-----------------------------------
Strongly disagree | 9 5.96 5.96
Disagree | 9 5.96 11.92
Somewhat disagree | 3 1.99 13.91
Somewhat agree | 15 9.93 23.84
Agree | 30 19.87 43.71
Strongly agree | 85 56.29 100.00
------------------+-----------------------------------
Total | 151 100.00

.

[Clyde Schechter]

The responses here form a 5-point Likert scale. If you buy the theory that Likert scales represent equally spaced responses on some latent numerical variable of agreement, then it is reasonable to assign a numeric value of 1 to Strongly Disagree, 2 to Somewhat Disagree, ..., 5 to Strongly agree and then average the numerical responses to the two questions. This would be consistent with a response model in which the response given to each question is a combination of "true" score and some random error. This is often done with survey items that are not exactly identical but have strongly overlapping content and where the separate questions are viewed as just minor variations in the wording (which sometimes makes a huge difference!). The case for doing it here, with an exact repetition of the question, is even stronger.