Standard Multiple Regression

Eleni A

New Member
#1
Hello,


Hypothetical scenario:

I’m designing a study which aims to exploring anxiety, depression, trauma in asylum seekers, and whether their ongoing stresses interact with anxiety, depression, and PTSD symptoms.

My hypothesis would be that those who experienced past trauma would be more vulnerable to the difficulties posed by the asylum application process.

I would collect demographic and socio-demographic variables and I would use 4 scales to measure anxiety, depression, and trauma, as well as collect data on their current difficulties (asylum application process, seeking for employment, accessing health-care, etc)

Would a STANDARD MULTIPLE REGRESSION be an appropriate statistical analysis to use for this design in order to explore relationships between the above mentioned variables?

Thank you in advance,
E
 

Eleni A

New Member
#5
Yeah, I am just asking you to write out the potential regression model.

difficulties = trauma + etc.
Thanks for trying help. I don't know how to answer your question. I think I got it completely wrong and confused in my head. Since I'm supposedly exploring relationships I should probably first conduct correlations between the variables - correlations between demographic and sociodemographic factors (pre-migratory trauma, delays in processing asylum application, racial discrimination, unemployment) and symptoms of anxiety, depression and PTSD, AND THEN conduct regression analyses to add ongoing stressors into regression models controlling for past traumas?

Please let me know if this makes more sense with the above design, and maybe advice me on which correlational test I should use?

E :)
 
#6
Can you define something like the following?

The dependent variables (Yi) :
anxiety - ordinal (1,2,3,4)
depression- ordinal (1,2,3,4)

Independent variables (Xi):
demographic - continuous
socio-demographic - continuous

Do you assume a linear connection between the dependent variables and the independent variables?
 

Eleni A

New Member
#7
Can you define something like the following?

The dependent variables (Yi) :
anxiety - ordinal (1,2,3,4)
depression- ordinal (1,2,3,4)

Independent variables (Xi):
demographic - continuous
socio-demographic - continuous

Do you assume a linear connection between the dependent variables and the independent variables?
I don't know to be honest. I just have this design in mind, and I don't know which statistical test I should use. I will go through my data and come back with an answer. Thank you for trying to help.
 

Eleni A

New Member
#9
Hey,

Apologies, my knowledge in Stats is so poor, I've tried to figure it out.

The scales which I'm going to use are all ordinal/ categorical. Since I will explore relationships, shouldn't I use a logistic regression instead as my "dependent" variables (anxiety, depression, trauma symptoms) will be categorical data. My predictor variables will be socioeconomic/ demographic characteristics, as well as pre-migration events and post-migration stressors. (These can be either nominal or categorical when using a logistic regression). Please enlighten me! I hope it makes sense now, and once again thank you! :)

E
 
#10
Hi Eleni,

if you use a Multiple Linear Regression you may get results smaller than 1 or bigger than 4.
You may probably use Multiple Linear Regression if you use it over an average of several questions, so it will become almost continuous dependent variable.

You can you Multinomial Logistic Regression for categorical data, But since your dependent variables are not just simple categorical variables (like red, blue, green), but ordinal data (1,2,3,4), you should get probably similar but a bit better results with the Ordinal Logistic Regression.

Since logistic regression is a bit confusing, you may use the following "Multinomial Logistic Regression" calculator for interpretation:
http://www.statskingdom.com/430logistic_regression.html
but probably better use ordinal Ordinal Logistic Regression for the final results

PS, are you independent variables continuous?
 
#11
Hello there,

Although this is an exploratory study because I'm looking for relationships, they are all categorical/ ordinal. For example age, gender, country of origin, and the additional checklist which will measure "post-migration stressors" will have a 5-point Likert Scale. I hope that answers your question. I read in my notes that with logistic regression, your IV can be either numerical or categorical.

Thank you for your great help!
 
#12
Correct, if categorical you need to use binary variables. You should read about using Likert Scale as categorical or numerical