Hello all,
I'm a beginner researcher (with 0 biostats experience) looking at risk factors that influence whether patients show up for their appointments or not. I'm trying to run a binomial logistic regression, but am currently stumped trying to decide which factors (about 15) to include in my model. Do I just include every independent variable in my regression, or is there an actual method to determine which groups of variables I should analyze at any given time? Thanks in advance!
You want to build a predictive model so there is an answer. The paper I am attaching has a protocol for doing what you want to do. If you have questions, please feel free to ask. Best wishes, David Booth
You should add all the 15 factors, but first you may want to check and record their influences at univariate level first, by running, one at a time, each factor against your dependent variable.
This way, you can see how they perform individually (univariate) and in group (multivariate) to appreciate the extent of confounding.
Your final prediction model will include only the significant variables seen in the multiple logistic regression.
However, there’s a lot data cleaning such as recoding variables (if you’re using SPSS) required. I’m happy to answer specific questions.
First, enter all of the variable in the model. Then remove those that are not meaningful. Next, obtain a logistic regression model with meaningful variable.
There should be large sample size. Again, multicollinearity and outliers must be absent in the data set. Selection of independent variables is crucial and one has to assess the fit of the model. Independent
variables with P value less than 0.25 in simple regression model should be included in the final model.
Please could you provide a literature reference regarding the choice of predictors based on the p value less than 0.25 (or 0.2)?
Thank you in advance
Purposeful selection of variables in logistic regression.
Bursac Z, Gauss CH, Williams DK, Hosmer DW.
Source Code Biol Med. 2008 Dec 16;3:17. doi: 10.1186/1751-0473-3-17.
As you could probably tell from the other responses, there is not a single way to select variables. Although you could certainly adhere to some sort of arbitrary threshold of significance from univariate analysis, sometimes it is best to include all of the non-collinear variables that you feel would be important and move from there.
You have to use all factors in your model .There are many ways that you can use to select your factors Y such as backward and step-wise selection method
in my experience, selection of variables depends on circumstances. for example if you got more variables with p
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Regression