Conflicting Signs in Correlation and Regression Analyses

[Serena Menny]

Hello,

When performing the correlation matrix, I found a positive and significant correlation between my dependent variable (sustainable investing) and the independent variable (sustainable concerns). However, when I included additional predictors in a regression analysis, the association reversed to negative and significant.

Furthermore, in a regression where the dependent variable is the intention to invest sustainably, the association with sustainable concerns remains positive and significant. Yet, when the dependent variable shifts to actual sustainable investing, the association becomes negative.

This pattern indicate a suppression effect?

Thank you,

[George Ford]

Yup.

"It is always with the best intentions that the worst work is done". Oscar Wilde.

[Serena Menny]

Yup.

"It is always with the best intentions that the worst work is done". Oscar Wilde.

okay thank youuu

[George Ford]

is this time series or cross sectional data?

[Serena Menny]

is this time series or cross sectional data?

cross sectional data

[Dimitriy V. Masterov]

Asked and answered here: https://www.statalist.org/forums/for...ounder-effects

[George Ford]

I would think this process would have a time dimension (concerned now -> intention now -> invest later). A snapshot in time can't detect such dynamics.

[Serena Menny]

I would think this process would have a time dimension (concerned now -> intention now -> invest later). A snapshot in time can't detect such dynamics.

Yes, this is an interesting insight. Thank you for providing it!

[Carlo Lazzaro]

Serena;
as an aside to previous helpful advice:
1) you compared different specifcations: therefore, no wonder that results changed;
2) -estat vce- shows the variance-covariance matrix of regression coefficients (which are not variables);
3) as per FAQ, sharing via CODE delimiters what you typed and what Stata gave you back is much more helpful for interested listers that tons of words aimed at describing your queries (all in all, statistics is a quantitative matter !). Thank you.

[Serena Menny]

Serena;
as an aside to previous helpful advice:
1) you compared different specifcations: therefore, no wonder that results changed;
2) -estat vce- shows the variance-covariance matrix of regression coefficients (which are not variables);
3) as per FAQ, sharing via CODE delimiters what you typed and what Stata gave you back is much more helpful for interested listers that tons of words aimed at describing your queries (all in all, statistics is a quantitative matter !). Thank you.

Thank you for your answer and your tips!
Kind regards,