3 replies to this topic

[mohsamir1984]

Dear All, 

In a disease model, I have infected and non infected. I have also for all these collected risk factor (age, gender, educational level, etc.) Using Binary logistic regression, I can calculate the contribution of each factor in causing the disease. Now, I want to calculate the contribution of risk factor combinations (i.e. if age + gender is more important or age + educational level ? ) and also in these combinations, how much is the contribution of each factor. So let us assume that Age + gender scored the highest contribution, but which one contribute more for the occurrence of the disease ? 

 

Thanks a lot 

[bob1]

What about a principal component analysis?

[mohsamir1984]

Good idea,, but it will only visually demonstrate which risk factor (variable) is more related to either infected or non infected groups. I have two questions: 

1. How one can express the contribution of each of risk factor ? i.e. which numerical value to be taken in this ? 

2. How PCA  can show the combination of risk factor (do you mean just visually ? ) can one express this as number ? 

 

Thanks  

[bob1]

I'm no expert, but my understanding was the PCA gave you values of percentage contribution of each variable, which you can then express as a plot if you want to. I don't know whether it can do combinations of the variables or not.