# Thread: Concerns with using a mix of Continuous and Categorical IVs in Logistic Regression?

1. ## Concerns with using a mix of Continuous and Categorical IVs in Logistic Regression?

Before I get into it, I want to make clear that I'm not a statistics guy. I'm doing what I can to get through a thesis and stumbling my way forward.

I'm using binary logistic regression. I have several independent variables most of which are continuous, but one of which is categorical (2 possible outcomes, yes or no). I'm to the stage where readers are reviewing my work and one of them asked, "What concerns are there with using IVs that are a mix between categorical and continuous?". I didn't know so I started looking around the internet. Either I'm not finding it or its right there and I just don't understand what I'm looking at, but after several hours of research, I have no idea what the answer to his question would be. Can anyone point me in the right direction?

2. ## Re: Concerns with using a mix of Continuous and Categorical IVs in Logistic Regressio

Hello,
from my (limited) experience with LR, I see no problem in having mixed IVs. I have seen many application of LR in research fields where both continuous and categorical predictors are used.

As for LR in general, I found this book interesting:
1. Allison PD. Logistic Regression Using the SAS System: Theory and Application. Cary: Wiley-Blackwell; 2001.
1. Hosmer DW, Lemeshow S, Sturdivant R. Applied Logistic Regression. Third. Hoboken, NJ, USA: John Wiley & Sons, Inc.; 2000.

More understandable to persons with not extensive math knowledge, I found this useful:
Field, Discovering Statistics using IBM SPSS Statistics

hope this helps
gm

3. ## Re: Concerns with using a mix of Continuous and Categorical IVs in Logistic Regressio

There are not issues with using continuous and/or categorical variables in a Logistic Regression model. It is designed to handle both types of variables.

The only things you need to be concerned with would be the proper use of them and interpretation. Given the program you used, you may need to define in the model if a variable is categorical and which category group is the reference group. Lastly, continuous and categorical variables have slightly different interpretations. So you need to make sure you are not mis-interpreting your results.

4. ## Re: Concerns with using a mix of Continuous and Categorical IVs in Logistic Regressio

Thanks!

hlsmith, could you expand on the different interpretations of continuous and categorical variables? I defined them properly using my software, but I'm not sure what the differences would be. Can I compare the odds ratios of different variable types against each other to determine which is more predictive?

5. ## Re: Concerns with using a mix of Continuous and Categorical IVs in Logistic Regressio

Continuous coefficients = log odds of a one unit increase in the variable, which you can find the exponential and get the odds ratio of X+1 over X, more or less. With X = the continuous variable. You can change the number of unit increase from one to whatever you desire.

Categorical coefficients = log odds of the variable versus the reference group. You can find the exponential of this coefficient and get the OR of the group versus the reference group.

So if you provide a description of your variables we can help you write out what you actually have.

6. ## Re: Concerns with using a mix of Continuous and Categorical IVs in Logistic Regressio

I see the difference. You are saving my life!

So, my study is on the impact of luck on the retention of the head coach in the NFL. My DV is whether the coach was fired or not. My continuous IVs are winning percentage, pythagoren win expectation (a predicted win percentage based on points scored and allowed), the last 5 year's winning percentage, the tenure of the head coach, and the number of years in a row that a team has improved or declined. The categorical IV is whether or not a team made the play-offs (1 if they did, 0 if they didn't).

7. ## Re: Concerns with using a mix of Continuous and Categorical IVs in Logistic Regressio

Many of these variables seem like they are saying the same thing. One needs to look out for collinearity of the IV. A result is large standard errors, which can make large effects not significant. Are all of your predictors significant? Have you thought about the relationship between the variables?

8. ## Re: Concerns with using a mix of Continuous and Categorical IVs in Logistic Regressio

I have been through all that with my adviser. All the predictors are significant except for the trend variable (how many years in a row a team has improved or decline). I tested for collinearity and, while many are in the same vein (ie performance of the team), they measure different enough aspects to not be collinear. I have been through many different models to get to where I am and none of the readers have a problem with the current one so I'm pressing forward with it.

 Tweet

#### Posting Permissions

• You may not post new threads
• You may not post replies
• You may not post attachments
• You may not edit your posts