I ran an analysis using the 2005 healthy eating index (HEI) score as the dependent variable and am a bit confused by my results. The HEI score ranges from 0 to 100, and categorized as bad (<51), needs improvement (52-80), and good (81-100). In my dataset, none of the participants scored in the "good" category, so I created a binary variable for my dependent variable. In the logistic regression models, the overall model is significant, and my primary independent variable of interest was also significant. I received a comment questioning my used of a binary variable, because dichotomizing variables typically results in losing valuable information. So, I ran my analysis again with the dependent variable as a continuous variable, and the overall model was not significant and the independent variable of interest was significant.

Am I doing something wrong? Why would I see significance with a binary versus continuous variable? Also, is there any reason to justify dichotomizing a variable? I thought it would be best to treat the dependent variable using its most practical form, and I have seen other precedence for this as well. Any help would be much appreciated. Thanks! ]]>

I'm trying to run a chi square test to compare two different screening methods for a pathogen. I'm testing this by diluting a known positive in a 10-fold dilution series and running those through two different screening methods. The way I'm setting up the chi square test is by using replicates as rows (3 reps, 3 rows) and the methods as columns (method 1, method 2). So I have 3 rows and 2 columns. For the observed values I'm using the endpoint detection (ex: if I ran 5 dilutions and was able to detect pathogen using one method as far as the 4th dilution in the first replicate then I used the number 4 as the observed value for that cell).

Does this make sense? Is there a better way to set up the table? The only other way I can think of to do it would be to use the number of times each method picks up sample at a certain dilution, but I'm not really sure how I'd organize that in the table.

I hope that makes sense. ]]>

I have some basic questions regarding logistic regression that I hope you can help me out with.

I’m running a simple logistic regression - output is attached. I need to calculate how much odds of smoking (smoker) is expected to fall/increase by when age increases by 10 years. I’ve read that it can be calculated in the following way:

0,99326^10 = 0,9346

Could you briefly explain if this is the correct method? Or should I multiply it with the constant - if yes why?

I’m also asked to calculate the odds of smoking of a 70-year old person when all x-variables are 0. Is it like this and why?

0,467*0,99326^70 = 0,29?

Your help is greatly appreciated. Thanks.

Jeff