0

I have a excel sheet with 100 independent variables and 1 dependent variable Now I would like to run stepwise regression using SPSS with it considering pairwise interaction between them to.

As it may be difficult to create all interaction columns manually

How to run a stepwise regression with interaction it considering

Any help as I am new to spss

Kind help please

Stephan Kolassa
  • 123,354
sriram
  • 59
  • 1
    Are you doing this for prediction, or for inference? – Stephan Kolassa Dec 06 '22 at 12:22
  • 2
    Why do you want to use stepwise regression? It's generally not the best approach. – Tim Dec 06 '22 at 12:38
  • 1
    Why would you like to run stepwise regression? It has a problematic reputation as something that produces rather spurious models that are difficult to interpret due to the instability of what terms are included in the model + the difficulty to do any inference post-mode-selection. – Björn Dec 06 '22 at 12:39
  • 1
    You don't. Algorithms for automatic model selection; Sane stepwise regression?; How to do stepwise regression with a binary dependent variable?; search the site for others. – gung - Reinstate Monica Dec 06 '22 at 12:40
  • 1
    Stepwise regression is a terrible idea in simple no-interaction settings. When considering interactions it's a total disaster. Think about the problem. Pre-specify models to the extent possible or use data reduction (unsupervised learning) first as discussed in my course notes chapter 4. – Frank Harrell Dec 06 '22 at 13:08
  • In addition to the comments about the issues with stepwise regression (which left out my favorite comment on the subject by Alexis), including all of the interactions means that you create an additional $(100(100-1))/2=4950$ interaction features. Do you really want to consider five-thousand features? – Dave Dec 06 '22 at 15:18
  • Here for me I have Labeled dependent variables as Y for a set of data. And a set of independent variables called features mostly to say. Now I need to pick the best features which could suit to find my dependent variable Y labeled. But I may need some transformations to get a better fit may be. Which could a better approach as i use spss. what techniques. – sriram Dec 06 '22 at 22:11
  • One more doubts If my dependent variable is Y and independents are X1,X2,X3 etc. Is it a good idea to get the fit of Log10(Y) in some terms of log10(X1), log10(X2) as a logarithmic fit will it tell the be way to predict values of log10(Y) – sriram Dec 07 '22 at 00:43
  • How many rows of data do you have? Are your Y values all positive so that a logarithmic transformation is even possible? Are your independent variables continuous or categorical? Please provide that information by editing the question, as comments are easy to overlook and can be deleted. – EdM Dec 08 '22 at 21:37
  • One help which could be the best Data transformations to be used to improve my F value in multiple linear regression as even though my R^2, Adjust R^2 are good and p value ok my F value only goes to 12 I have tried Reciprocal, square, Square root , Normalize etc any other suggestions please – sriram Dec 09 '22 at 14:22
  • It's not clear why you are so worried about the value of your F statistic if other measures of model fit are OK. The problem pointed out in several comments, if you built the model by stepwise selection, is that all of those values are likely to be highly optimistic and unlikely to be valid on new data samples. We have over 2000 pages on this site with the data-transformation tag; you haven't provided enough information about your data and the model in the question to get any more specific suggestion. – EdM Dec 10 '22 at 16:02

0 Answers0