Can I use compute variable to club my different dependent variables into a single dependent variable and go for linear regression analysis in SPSS? I want to know if this is allowed in data analysis and which citation can I refer to?
1. Why you want to merge several independent random variables to get one random variables? What is your purpose?
2. How you can develop the linear regression model without dependent random variable?
3. The linear regression model/analysis is a technique used to predict the value of one quantitative variable by using its relationship with one or more additional quantitative variables. For example, if we know the relationship between height and weight in adult. males, we can use regression analysis to predict weight given a particular value for height.
4. SPSS can be used easily to estimate the linear simple/multiple regression model. Please refer to:
. Landau, S. and Everitt, B. S.(2004). A Handbook of Statistical analyses using SPSS.
.Howitt, D. and Cramer, D.(2008). Introduction to SPSS.
Yes, I have understood your point Zuhair Sir and Mr. Firdos Khan. But my intention is to create a new dependent variable from many similar related dependent variable. Say for example: I want to measure technology related banking services where I have included ATM, Mobile Banking, Internet Banking, Wallets, as dependent variables and also solicited responses for using these services. My intention now is to club these variables into single dependent variable so that I can run a regression model with several independent variables that I have.
COMPUTE variable can be used to apply many transformations related to variables. The way you are seeking can be easily applied by taking an average of all the variables. However, the question is whether such combination of variables can be justified theoretically & logically.
It may be that you have taken the above mentioned 'variables' as an item to measure a 'Factor/Variable' or are you trying to create a second order factor from the existing variable? This needs to be understood.
If you have several highly correlated variables (with the same sign) then taking a simple average and using them in a regression is a perfectly reasonable thing to do. It is an atheoretical way of combining variables but one that it is transparent and will be fairly robust.
You can also use formal methods to combine variables. Some (such as MANOVA) in my view are worse than simple averaging - because they capitalise on chance, lack transparency and so forth. Others such as principal components are worth considering but if you have a set of k highly correlated variables then PCA and simple averaging will almost certainly produce very similar results. With more complex patterns of variation PCA would be more useful or interesting.
Note that all three approaches are equally atheoretical but simple averaging is probably the easiest to produce a theoretical justification for.
When you say "I have included ATM, Mobile Banking, Internet Banking, Wallets, as dependent variables," you mean you have multivariate regression, and you are using the same independent variables in each case for multivariate multiple regression? I wasn't sure if you were being careful when saying what was "independent," and what was "dependent." At first I thought you meant you wanted to combine independent variables, to "...go for linear regression analysis...," using y = y*+e, if y* is predicted y.
But now it sounds like you do mean to take the various regressions, and average results. Not sure why. But if you do, I'd think you'd want a weighted average, based on frequency of each banking method that you listed. ???
Cheers - Jim
PS - If you just ignore the difference between dependent variables, and use only one multiple regression, won't that automatically average them?
PSS - I am not familiar with today's SPSS, or "club," or "COMPUTE" commands.