Problems with Variable Exclusion in Multiple Regression

[Thommy]

Dear everybody,

I have conducted a survey where people were first asked about their general perceptions  and the importance of justice and ethics.

Then later, there were four different scenarios of a company behaving in different ways (neutral, negative, positive, mixed) which were randomly assigned, so each person only read one scenario. Then they had to answer questions about their perception of the situation as ethical or unethical.  (all Likert scales from 1 to 7).

Now I have conducted ANOVAs to see if there are significant differences in mean of perception score depending on which scenario the participants read. After that I conducted an independent samples T-Test to evaluate whether the mean of the perception scores were higher for people that had read the positive scenario than the negative one (which they were).

Now I am looking at a possible moderation, whether the general perception of justice has an impact on how people that read the case of the firm behaving positively judge the situation as ethical or unethical. The same I am doing for the negative scenario.

I have dummy coded the two scenarios "positive" and "negative" and calculated the average scores for the outcome variable for the positive scenario. Additionally, I calculated an interaction term for the IVxModerator.

The problem I now have is: SPSS excludes all except for the moderator term from the regression. Can that be due to the fact that I am using the average perception scores for the people with the positive scenario that were and also the average scores for the justice perception from the beginning also for these people ? The error code is:

"For models with dependent variable pCSRMP, the following variables are constants or have missing correlations: Zscore(PositiveU). They will be deleted from the analysis."

Could that be due to multicollinearity ?

The only IV left in the regression is the interaction term. Is that normal ?

Thank you all so much for your help

Best regards
Thommy

[Bruce Weaver]

Thommy, please post your syntax for computing new variables (i.e., indicator variables & product terms) and for estimating the regression model.  If you did not use syntax, you can copy the relevant commands from the output window (or the journal file).  Thanks.

Thommy wrote

Dear everybody,

I have conducted a survey where people were first asked about their general perceptions  and the importance of justice and ethics.

Then later, there were four different scenarios of a company behaving in different ways (neutral, negative, positive, mixed) which were randomly assigned, so each person only read one scenario. Then they had to answer questions about their perception of the situation as ethical or unethical.  (all Likert scales from 1 to 7).

Now I have conducted ANOVAs to see if there are significant differences in mean of perception score depending on which scenario the participants read. After that I conducted an independent samples T-Test to evaluate whether the mean of the perception scores were higher for people that had read the positive scenario than the negative one (which they were).

Now I am looking at a possible moderation, whether the general perception of justice has an impact on how people that read the case of the firm behaving positively judge the situation as ethical or unethical. The same I am doing for the negative scenario.

I have dummy coded the two scenarios "positive" and "negative" and calculated the average scores for the outcome variable for the positive scenario. Additionally, I calculated an interaction term for the IVxModerator.

The problem I now have is: SPSS excludes all except for the moderator term from the regression. Can that be due to the fact that I am using the average perception scores for the people with the positive scenario that were and also the average scores for the justice perception from the beginning also for these people ? The error code is:

"For models with dependent variable pCSRMP, the following variables are constants or have missing correlations: Zscore(PositiveU). They will be deleted from the analysis."

Could that be due to multicollinearity ?

The only IV left in the regression is the interaction term. Is that normal ?

Thank you all so much for your help

Best regards
Thommy

[David Marso]

Also consider doing the obvious and run correlations descriptives and frequencies to verify the warning/error message.

Bruce Weaver wrote

Thommy, please post your syntax for computing new variables (i.e., indicator variables & product terms) and for estimating the regression model.  If you did not use syntax, you can copy the relevant commands from the output window (or the journal file).  Thanks.

Thommy wrote

Dear everybody,

I have conducted a survey where people were first asked about their general perceptions  and the importance of justice and ethics.

Then later, there were four different scenarios of a company behaving in different ways (neutral, negative, positive, mixed) which were randomly assigned, so each person only read one scenario. Then they had to answer questions about their perception of the situation as ethical or unethical.  (all Likert scales from 1 to 7).

Now I have conducted ANOVAs to see if there are significant differences in mean of perception score depending on which scenario the participants read. After that I conducted an independent samples T-Test to evaluate whether the mean of the perception scores were higher for people that had read the positive scenario than the negative one (which they were).

Now I am looking at a possible moderation, whether the general perception of justice has an impact on how people that read the case of the firm behaving positively judge the situation as ethical or unethical. The same I am doing for the negative scenario.

I have dummy coded the two scenarios "positive" and "negative" and calculated the average scores for the outcome variable for the positive scenario. Additionally, I calculated an interaction term for the IVxModerator.

The problem I now have is: SPSS excludes all except for the moderator term from the regression. Can that be due to the fact that I am using the average perception scores for the people with the positive scenario that were and also the average scores for the justice perception from the beginning also for these people ? The error code is:

"For models with dependent variable pCSRMP, the following variables are constants or have missing correlations: Zscore(PositiveU). They will be deleted from the analysis."

Could that be due to multicollinearity ?

The only IV left in the regression is the interaction term. Is that normal ?

Thank you all so much for your help

Best regards
Thommy