I wonder how to model an interval variable using linear regression. My independant variables contain both continuous, dummy, ordinal and interval variables!
If your dependent variable (DV) is classified with interval, such classification can be transformed as 1, 2, 3, . . . as an ordinal variable OR get the middle point of the classes (like if the classes are 1-10, 11-20, . . . , then the middle points are 5, 15 and so on, if the classes 0-
If your DV is interval and continuous, then you likely are able to use ordinary multiple linear regression. Categorical and ordinal variables can be re-expressed as dummy variates to be included as IVs.
If there's something else unique about your study that might be germane, you'd likely get more useful recommendations if you were to elaborate on those features.
Soukaina Aziz I am going to disagree a bit with my friend Dr . David Morse This question came up here recently and several people dug into it and I don't believe you want to treat a bounded continuous DV with standard multiple regression. As you will see in the attached thread there are a number of better choices that depend on the exact conditions of your DV. I hope that Dr. Morse will comment on these. Here's the link: https://www.researchgate.net/post/Tell_the_distribuition_of_this_dataset_Gamma_distribution
Can you say more what you mean by linear model? If you want a linear model, you just do that. There are some assumptions about things like calculating the parameters and associated statistics (e.g., standard errors), but if your model is linear between the IVs and the DV, then IT IS. What you do with that model and how you estimate it may require some assumptions, but without further of information I can't suggest much.
Daniel Wright it's not about hiding information! it's only about sharing files which is more practical by gm. I will try to resume my demand and post a question! hope you could help
Thank you for your answers and sorry to give a delayed answer
My dependent variable is a continuous interval:
0
20-30
30-50
More than 50
I made assumptions to restrict (More than 50) interval and it becomes 50-300
I tried it to model it as a linear regression considering the midpoint of each variable but it doesn't work! I also tried ordinal variable in SPSS
My independents variables are: ordinal variables (I turned them into dummy ones using a number for each var) and binary variable
Even if my variables are not correlated, my IV don't explain variability of my DV via regression! But when analysing data it's clear that lot of IV influence the DV
It looks as if you do not have a continous dependent variable. If I understand this correctly, you have 4 categories (what about values between 1 and 19, or is this just a typo?), which are ordered, which suggests an ordinal regression. But please show us the results you are talking about. It remains unclear how your results look like and what may cause the problem. Correlations, scatterplots and the regression results would help.