# Compare regression coëfficiënts in 2 models/ same sample, + 1 IV, SPSS

#### Mathis

##### New Member
Dear all,

I am doing a master thesis in criminology and am trying to detect confounding effects of response styles on theoretical models. Therefore I wanted to bring in this confounding variable in a theoretical model and see how his affects the model. I have searched around but found either solutions for programs I do not have access to or they are not applicable to this specific situation.

I perform the following regression analyses (negative binomial due to variable specifications):

X = b1Y + b2Z
X = b3Y + b4Z + b5C

where
X= number of crimes in past year
Y= delinquency of friends
Z= parental control
C= counfounding factor (tendency of respondent to answer ambiguous, neutral)
- the same sample is used in both models

My hypothesis is that people who answer more neutrally (presumably because they have no interest in the research and do no effort to formulate an opinion) confound the variables and thus the relationships. If the variance explained by this confounder is controlled for by adding it to the model and coefficients b1 and b2 change significantly, this response style does cause problems. If it does not change significantly, there is no problem.

I did the test and my coefficients were:
b1 = .445*** | b3 = .443***
b2 = -,103 | b4 = -.105
b5 = -.015

Although at first sight you can see adding C had little to no influence, I am going to do the same for other confounders and other DV's where a greater change may occur so I need a standard method to say when it changes and when it does not.

ps: I work with SPSS
ps': I cannot do SEM, which is apparently recommended.

Kind regards,
Mathis

#### Mathis

##### New Member
Hello,

I think I will use the following formula:

Z= b1 - b2 / sqr(SEb1²+SEb2²)

source: Clogg, Clifford C., Eva Petkova, and Adamantios Haritou
Statistical methods for comparing regression coefficients between models.
American Journal of Sociology 100:1261-1293.

#### Dason

Hello,

I think I will use the following formula:

Z= b1 - b2 / sqr(SEb1²+SEb2²)

source: Clogg, Clifford C., Eva Petkova, and Adamantios Haritou
Statistical methods for comparing regression coefficients between models.
American Journal of Sociology 100:1261-1293.