Take the 2-minute tour ×
Cross Validated is a question and answer site for people interested in statistics, machine learning, data analysis, data mining, and data visualization. It's 100% free, no registration required.

If I want to use some proportion type independent variables in a logistic regression, then what will be the interpretation of the regression coefficients corresponding to those proportion type variables? Will that mean- "The change in log odds for per unit change in the proportions"?

But what will be meant by "per unit change" in this case? As the proportions lie within [0,1], I am getting a little confused with what a "per unit change" will mean in this scale. Does it mean 0.01 or 1%? (I am sorry for my noob thoughts!)

In my data the range of the proportions is 0 to 1, not multiplied by 100. Do I need to multiply them by 100? So that I can say "per unit change" means 1% change? I have seen that the coefficients do differ in scale if I multiply the proportions by 100. For example, a coefficient of -1.3 for proportions becomes -0.013 for percentages (when the proportions are multiplied by 100).

share|improve this question

2 Answers 2

up vote 1 down vote accepted

The interpretation for the regression coefficient is always for a 1 unit change regardless of what a "unit" is. In your case, if the IV is a proportion falling between 0 and 1, a one unit change is the same as 100%.

If instead you want to look at the "effect" of a 1% change, simply multiply your IV by 100 before using it in the regression.

share|improve this answer
    
Thank you. You are absolutely right. Actually FMZ is also right. "Per unit change" for a proportion type variable should always mean change by 1 (or 100%) unit in the independent variable. Just that for my example, Y goes down by 1.3 units when X (measured in proportions) increases by 1 unit also implies that, Y goes down by 0.013 (1.3/100=-0.013) units when X (measured in proportions) increases by 0.01 (1/100=0.01) units. That makes it consistent with the sense of proportions. Thank you all. –  Blain Waan Jan 11 '13 at 11:16

They are the same, aren't they?

Let's take the first model, coefficient is -1.3 and variable is in original scale (0-1). So if the variable increases by 0.01, say from 0.08 to 0.09, then the log odds are down by 1.3*0.01 = 0.013.

Now the second model, coefficient is -0.013 and variable is multiplied by 100 (0 - 100). So if the variable increases by 1 from 8 to 9 (which is actually from 0.08 to 0.09), then the log odds are down by 0.013*1=0.013.

share|improve this answer
    
Dear FMZ, thank you for your kind reply, but could you please confirm me when I use proportions as independent variables then "per unit change" in the independent variable means an increase by 0.01. I was guessing it, but as it is a continuous variable just like the other continuous variables (which would generally mean an increase by 1), so I was afraid if a unit change meant '+1' or '+0.01' in this case! Probably very immature question, but I would like you to explain this a little. –  Blain Waan Jan 11 '13 at 10:15
    
Ok I have got it. I would like to thank you too for your kind thoughts. –  Blain Waan Jan 11 '13 at 11:21

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.