Hi,
I'm new to Jamovi (and to statistics in general, to be honest), and I need to analyze my data to see whether certain factors contribute to a selection. My dependent has 4 levels, and I have factors that have 2 or more levels, such as gender (male, female), hunger (low, high), and BMI (underweight, normal, overweight, obese). So far, I have done the analysis by following tutorials, but I have questions regarding interpretation.
In all tutorials I have seen, the predictors are continuous, and I understand how to interpret that. However, I have trouble interpreting the factors.
Looking at the screenshot from my analysis, the only 2 significant p values are for underweightnormal and obesenormal. Considering these, could you please help me interpret the direction of this effect?
Thank you so much,
Beth
Multinomial Logistic Regression Interpretation
Multinomial Logistic Regression Interpretation
 Attachments

 multinomial.PNG (29.22 KiB) Viewed 462 times
 mcfanda@gmail.com
 Posts: 391
 Joined: Thu Mar 23, 2017 9:24 pm
Re: Multinomial Logistic Regression Interpretation
Hi,
you may want to have a look at Agresti (2007), https://mregresion.files.wordpress.com/2012/08/agrestiintroductiontocategoricaldata.pdf
In a nutshell, multinomial predicts comparisons between one group defined by the dependent variable and every other group using a series of logistic regressions. You categorical independent variables (IV) are also cast in the model as dummy variables. Thus the coefficients are the difference in IV groups in the (logit) of being in one group of the dependent rather then the other.
More practically, the exp(B) associated with underweightnormal tells you the ratio between the odd of being "processed low cal" vs "natural high cal" between underweight and normal. Being practically 0 (109e10), it means that underweight group odd to be "processed low cal" is practically largely smaller than the odd for normal group.
To make it easier to read, I'd suggest to change the order of the levels in the dependent variable (you can do that in the data tab of jamovi), setting processed low cal as the first level.
you may want to have a look at Agresti (2007), https://mregresion.files.wordpress.com/2012/08/agrestiintroductiontocategoricaldata.pdf
In a nutshell, multinomial predicts comparisons between one group defined by the dependent variable and every other group using a series of logistic regressions. You categorical independent variables (IV) are also cast in the model as dummy variables. Thus the coefficients are the difference in IV groups in the (logit) of being in one group of the dependent rather then the other.
More practically, the exp(B) associated with underweightnormal tells you the ratio between the odd of being "processed low cal" vs "natural high cal" between underweight and normal. Being practically 0 (109e10), it means that underweight group odd to be "processed low cal" is practically largely smaller than the odd for normal group.
To make it easier to read, I'd suggest to change the order of the levels in the dependent variable (you can do that in the data tab of jamovi), setting processed low cal as the first level.
Re: Multinomial Logistic Regression Interpretation
Oh I see! Thank you so much for taking the time to explain, I understand much better now!
mcfanda@gmail.com wrote: ↑Fri Feb 03, 2023 8:49 am Hi,
you may want to have a look at Agresti (2007), https://mregresion.files.wordpress.com/2012/08/agrestiintroductiontocategoricaldata.pdf
In a nutshell, multinomial predicts comparisons between one group defined by the dependent variable and every other group using a series of logistic regressions. You categorical independent variables (IV) are also cast in the model as dummy variables. Thus the coefficients are the difference in IV groups in the (logit) of being in one group of the dependent rather then the other.
More practically, the exp(B) associated with underweightnormal tells you the ratio between the odd of being "processed low cal" vs "natural high cal" between underweight and normal. Being practically 0 (109e10), it means that underweight group odd to be "processed low cal" is practically largely smaller than the odd for normal group.
To make it easier to read, I'd suggest to change the order of the levels in the dependent variable (you can do that in the data tab of jamovi), setting processed low cal as the first level.