Goodmorning everyone,
I need help interpreting a logistic regression model. I'm trying to learn a bit more about statistics and the use of Jamovi. I'm considering a dataset about how some variables can influence the presence of High blood pressure (HBP). So, for example, I'm considering wether low/medium/high sodium intake can affect the risk of this pathology.
As well known sodium intake is a risk factor of HBD.
I tried to use Jamovi by setting my reference levels: hypertension yes, and sodium intake low.
As you can see from the attached files I would interpret my results as:
- medium sodium intake, compared to low sodium intake, increases the risk of HBP development (OR 1.12); by the way, since the CI 95% includes the value "1", I wouldn't consider to consider this value in a multiple logistic regression model;
- on the other side, high sodium intake vs low sodium intake seems to be a protective factor against HBP development (OR 0.11).
So I'm wondering if I'm making some mistakes in the interpretation of these values or if my dataset has data against what is known in the literature.
Thank you in advance for your kind help
Logistic model - help needed
Logistic model - help needed
- Attachments
-
- Jamovi pic.png (49.91 KiB) Viewed 46003 times
-
mcfanda@gmail.com
- Posts: 549
- Joined: Thu Mar 23, 2017 9:24 pm
Re: Logistic model - help needed
set `hypertention_status` `livello di riferimento` to `no`, and the results would make more sense. In this model, you're predicting the probability (the odd to be precise) of "not be hypertension_status=yes". I guess you want to do the opposite. You'll see that the "Test Globale del Modello" would not change, the "Stima" will change the sign, and the odd-ratio would be more in line with the literature (and my doctor saying to me to be in the low-sodium group 
)
)Re: Logistic model - help needed
Thank you very much!
Re: Logistic model - help needed
Now I'm trying to implement a multiple logistic regression model. The outcome is always hypertension yes/no and I'm considering a set of variables.
First of all I calculated the OR for each variables in a simple logistic regression model, so I identified the ones whose OR CI 95% did not include the value "1" (the ones followed by a
.
The variables are:
- age
- gender
- sodium intake (low-medium-high)
- physical acticity level (from 1 to 5)
- family history (yes/no)
- income level (low, lower middle, upper middle, high) *
- education (middle school or lower, high school, university degree or higher) *
Should I use all of these variables in my model? Or should I avoid to use the ones marked by a *?
First of all I calculated the OR for each variables in a simple logistic regression model, so I identified the ones whose OR CI 95% did not include the value "1" (the ones followed by a
.The variables are:
- age
- gender
- sodium intake (low-medium-high)
- physical acticity level (from 1 to 5)
- family history (yes/no)
- income level (low, lower middle, upper middle, high) *
- education (middle school or lower, high school, university degree or higher) *
Should I use all of these variables in my model? Or should I avoid to use the ones marked by a *?