I agree with Cristian Ramos-Vera , that more information as to what your dichotomous variable(s) are, how they were collected, and how many you have, as well as your specific research questions/aims is required in order to offer constructive advice. Would you kindly elaborate?
I have a question in my data that asks participants use of media as no and yes (1, and 2). Now I want to build a model where I need to convert this dichotomy into at least three or four levels in order to predict a multi-exogenous variable.
The Recode - Into Different Variables procedure is launched from the Transform menu.
The dichotomous variable is passed in the Numeric Variable - Output Variable zone. Fill in the name of the variable to be created by recoding in the Name area.
Press the Old and New Values button and describe the boundaries of the recoding classes. Then press the Continue - Ok button to register the new variable in SPSS.
Thank you Christian, David, Taqi, and Marius for your time, questions and possible answers.
Marius, what range should I assign to 1 and then which range would be appropriate for 2. For example if I want to change it to a 4 point scale can I recode 1 into 1 and 2, and 2 in 3 and four. Also would it be appropriate to recode this data especially if I am intending to use AMOS for SEM.
Gulnaz Anjum: I'm afraid that the answer to your original question is no. You can't transform dichotomous data into continuous data. It would require making very strong and untenable assumptions. You would need to make assumptions about the distribution of the variable, which could bias the results of your analysis. To illustrate, you cannot evenly divide people who answered "no" into 1 and 2 (i.e. if 60% answered no, then there is no reason to assume that 30% would answer "1" and 30% would answer "2", other distributions are also plausible, i.e. 45% and 15%). People reviewing your study would probably criticize such strong assumptions. If you need ordinal or metric data then you need to consider this when designing your study.
In this case the recoding of the variable is limited to the number of its values, respectively 1 and 2. I think in this case it is recommended to recode another demographic variable that has a greater distribution of values, respectively the age variable, city size, family type, etc. In the case of the age variable you can opt for example for 5 or 10 year classes expressed as follows: 29 thru 39 - 1; 40 thru 49 - 2 etc.
Regarding your variable I think it can be recoded in group 1 of those who do not use media and 2 in the group of those who use media. With the two groups you can make comparisons between them by including other variables with a larger distribution.
Splitting the database into subgroups can be done using the Data - Split File procedure.
Thank you Lukasz, I totally agree with you. I have done my analysis based on categorical variable and I was suggested to transform my variable. Thank you for your answer, I am very grateful for the reaffirmation.