hello everyone. i have multiple continuous dependent variables rated on scale along with categorical independent variables such as sex, study etc. while checking for linearity i realised that my data was non linear for most of the variables. can someone suggest to me how can i analyse my data in spss? i was planning to do pearson´s correlation, glm multivariate and multiple regression analysis but because i was too late to think of assumption i have to start over again. your help would be immensely appreciated.
so far i conducted cosine of vector of values as i read that it was suitable for non linear correlation. is it true? also while comparing pearson with cosine my correlation value for pearson matches the similarity test for cosine test. is this usual?
hoping for your kind and important help.
You might want to try the book in the attached screenshot.. it's available in the z-library. Best wishes David Booth
Hi Nisha Regmi
Performing non linear regression requires a primary knowledge about your function. There are many non-linear function to be used. You can try some of them like exponential, polynomial and etc and check the determination coefficient and mean squared error. You can use some program such as CureExpert Professional to find the function. I don't know any handbook for performing non-linear analysis in SPSS but i would suggest "NONLINEAR REGRESSION MODELING FOR ENGINEERING APPLICATIONS" for study.
Good luck
David Eugene Booth thank you very much. finished downloading and i will check it out.
@Milad Razbin thank you for your suggestion. indeed I don't know much about function as I'm not a physical science major. but I did take your suggestion and looked up the polynomial regression. I am trying to see if it will work for me. thank you for your help.
Please 🙏 look up the definition of nolinear regression. YOU will be very very surprised. In fact I'm attaching one for you to consider. Best wishes, David Booth
David Eugene Booth if i understand correctly linear regression is different from the non-linear model as it is based on the fitted approximation of the original data. is it correct? for me i find it hard to understand concept like this. so far, what i have understood is that linear relationship is where you can fit the data in a straight line which is different in non linear relationship and for the non linear regression you create a model that tries to linearise the variables first and then run the analysis. i'm still kind of confused. is there a better way to understand the differences between them in a simplest term? once again, extremely grateful for the help and support.
The take home message is you likely don't need this. Stick with several multiple regressions one for each dependant variables.
Best wishes, David Booth
- Badges
- Science topic
- Similar topics
- Analytical Chemistry
- Column