Hello,
I have a question about the multiple regression model. I want to fit a multiple regression model on a dataset in which the dependent variable is continuous and all of the independent variables are categorial. Should I check the assumption of the multiple regression model also for categorical variables? if they are not met, which model may I use instead of the multiple regression model to analyze the effects of these variables on the dependent variables?
thank you
Forget the assumptions, you can insert dummy (yes/no) binary bariables in a multiple regreesion model, clearly this holds true if and only if you have a yes/no structure, but even if you have, let's day, three categories you can split the three category variable into a three yes/no binary items, see:
https://www.jstor.org/stable/2346482
Article Quantitative Structure−Activity Relationships of Mutagenic a...
Use standard multiple regression do all of the checks as usual. Best wishes David Booth
Alessandro Giulliani, you can run a log linear regression with your data set using the R statistical package.
Well, if you have 1/0 (the most convenient choice for dummy variables) log linear is not applicable...
Yes, Alessandro Giulliani, the R statistical package will do the log linear regression for you. It is not available in SPSS yet.
That was the basis for suggesting your use of R statistical package.
Why is handling of dummy variables in your data set an issue ?
Dear Harold, it is an issue because the answer deals with the insertion of categorical variables ina mutiple regression together with continuous ones.
..clearly if all the independent variables are categorical you can shift to logistic
Get a book on regression. This a standard ANOVA type regression problem. I am attaching such a set of notes for anyone that is interested. David Booth
Just can make an invariable model,by the square model and next compare with one by the MML Regression.
If it is uncertain whether the influence of various factors on the results is linear or nonlinear, it is better to use a nonlinear model.
http://www.eaihua.com/drs-e.html
The assumtions should always be met (at least approximately), independent of the predictors. You can check this by looking at residual diagnostic plots.
If the assumption of the normal distribution is not reasonable for the response, you might consider a suitable transformation of the response or use a model with a more appropriate distribution model (generalized linear models offer some useful alternatives).
Koen Van de Moortel my dependent variable is numeric and all of the predictors are categorical, can I still use this test?
Alessandro Giuliani thank you for your comment, so, if the dependent variable is numeric and all of predictors are categorial , is it not needed to check linearity and other assumptions?
David Eugene Booth thank you for your comment, when all the predictors are categorial with many different levels, how can I check linearity and other assumptions?
Yanping Wang thank you for your comment, all of my predictors are categorial, and the dependent variable is a numeric very skewed one, how can I test the linearity? can i use log transformation for the dependent variable and run the multiple regression?
Jochen Wilhelm my predictors are categorial, and the dependent variable is a numeric very skewed one,can i use GLM model?
Samaneh H.Bahreini , yes, but better go and take help from a statistician and follow David's advice and read some books about regression models.
In my humble opinion, the main and secondary objectives of the research study are in terms of the type of variables (quantitative or qualitative), graphical examination of data, the determination techniques of using statistical methods in accordance with the desired assumptions (parametric and non-parametric), the type of probablistic relationship (correlation or cause and effect), the technique of using influential variables and the use of prevalent data reduction methods as basic prerequisites in using multiple analysis tests. Then, in the next steps, according to the being specialization of the subjects (mean or variance), other related tests should be used. In fact, The type of data and its initial investigation are very important for defining and determining any statistical test.
Sam, categorical predictor variables don't matter here. The DV determines the type of regression that you want to use. Best wishes David Booth
Yanping Wang my dependent variable is continuous, my predictors are categorial with different levels, my dataset has 8633 rows of the households,
Factors can affect multiple results of multiple aspects of a system. There may be different key factors. Each group of factors must have corresponding results. Otherwise, this is an incomplete sample and cannot be used in modeling.
Factors can affect multiple results of multiple aspects of a system. There may be different key factors. Each group of factors must have corresponding results. Otherwise, this is an incomplete sample and cannot be used in modeling.
Based on the descriptions of your variables, I think this article helps answer your questions. Article An Ethnic Advantage: Teacher-Student Ethnicity Matching and ...
Dear @Sam Benhein, if the independent predictor variables are not also continuous variables, the regression type is log linear regression.
- Badges
- Science method