[Quint Koevoets]

I have generated the following model with SEM at the moment:

sem (csp2012 -> csp2013) (csp2012 -> csrrep2012) (csp2012 -> csrrep2013) (csp2013 -> csp2014) (csp2013 -> csrrep2013) (csp2013 -> csrrep2014) (csp2014 ->
> csrrep2014) (csrrep2012 -> csrrep2013) (csrrep2012 -> roa2012) (csrrep2012 -> roa2013) (csrrep2013 -> csrrep2014) (csrrep2013 -> roa2013) (csrrep2013 ->
> roa2014) (csrrep2014 -> roa2014) (roa2012 -> roa2013) (roa2013 -> roa2014), nocapslatent
(47 observations with missing values excluded;
specify option 'method(mlmv)' to use all observations)

Endogenous variables

Observed: csp2013 csrrep2012 csrrep2013 csp2014 csrrep2014 roa2012 roa2013 roa2014

Exogenous variables

Observed: csp2012

Fitting target model:

Iteration 0: log likelihood = -1409.5922
Iteration 1: log likelihood = -1409.5922

Structural equation model Number of obs = 59
Estimation method = ml
Log likelihood = -1409.5922

---------------------------------------------------------------------------------
| OIM
| Coef. Std. Err. z P>|z| [95% Conf. Interval]
----------------+----------------------------------------------------------------
Structural |
csp2013 <- |
csp2012 | .8952065 .0538061 16.64 0.000 .7897486 1.000664
_cons | 4.812915 2.67275 1.80 0.072 -.4255786 10.05141
--------------+----------------------------------------------------------------
csrrep2012 <- |
csp2012 | .0293492 .0244935 1.20 0.231 -.0186571 .0773555
_cons | 66.6809 1.216683 54.81 0.000 64.29625 69.06556
--------------+----------------------------------------------------------------
csrrep2013 <- |
csp2013 | -.0384826 .0226693 -1.70 0.090 -.0829137 .0059485
csrrep2012 | .9794071 .0497989 19.67 0.000 .881803 1.077011
csp2012 | .0315363 .0222354 1.42 0.156 -.0120442 .0751168
_cons | -.2143459 3.342937 -0.06 0.949 -6.766383 6.337691
--------------+----------------------------------------------------------------
csp2014 <- |
csp2013 | .7961412 .0864879 9.21 0.000 .6266279 .9656544
_cons | 6.387988 4.27547 1.49 0.135 -1.991779 14.76775
--------------+----------------------------------------------------------------
csrrep2014 <- |
csp2013 | .0238833 .0139535 1.71 0.087 -.003465 .0512316
csrrep2013 | .830931 .0444756 18.68 0.000 .7437604 .9181015
csp2014 | -.0189221 .0133641 -1.42 0.157 -.0451154 .0072711
_cons | 11.23263 2.92878 3.84 0.000 5.49233 16.97294
--------------+----------------------------------------------------------------
roa2012 <- |
csrrep2012 | 1.034393 .3241771 3.19 0.001 .3990179 1.669769
_cons | -64.35061 22.08696 -2.91 0.004 -107.6403 -21.06097
--------------+----------------------------------------------------------------
roa2013 <- |
csrrep2012 | .0134556 .3444018 0.04 0.969 -.6615595 .6884706
csrrep2013 | .1792114 .3208322 0.56 0.576 -.4496082 .808031
roa2012 | .7536004 .0496173 15.19 0.000 .6563522 .8508485
_cons | -11.66387 8.920036 -1.31 0.191 -29.14682 5.819076
--------------+----------------------------------------------------------------
roa2014 <- |
csrrep2013 | .1854684 .3669558 0.51 0.613 -.5337517 .9046886
csrrep2014 | -.0118446 .4074981 -0.03 0.977 -.8105263 .786837
roa2013 | .588486 .0683561 8.61 0.000 .4545104 .7224615
_cons | -8.877053 10.91728 -0.81 0.416 -30.27453 12.52043
----------------+----------------------------------------------------------------
Variance |
e.csp2013 | 31.32861 5.768065 21.83847 44.94279
e.csrrep2012 | 6.492016 1.195277 4.525439 9.313191
e.csrrep2013 | .9390427 .1728918 .6545856 1.347114
e.csp2014 | 78.69509 14.48894 54.85658 112.8929
e.csrrep2014 | .8241646 .151741 .5745067 1.182314
e.roa2012 | 41.2324 7.5915 28.74218 59.15038
e.roa2013 | 5.868479 1.080475 4.090785 8.41869
e.roa2014 | 8.305627 1.52919 5.789666 11.91493
---------------------------------------------------------------------------------
LR test of model vs. saturated: chi2(20) = 66.25, Prob > chi2 = 0.0000

I would like to ask where I can add my covariates? For example if you have the visual version of the model, can I add a row with 3 observed (or latent) variables that are under y1 y2 and y3 and have paths with y1 y2 and y3?

[Quint Koevoets]

I have generated the model by building it with SEM. I have added the model in the attachment. I would like to ask where I can fit in my covariate variable rdexpenses? Since it is not an extra wave I assume it can be put under y1 y2 y3 and make paths for example like: rd1 -> y1 rd1>y2 rd1>rd2 etc.? And are the covariates latent or observed variables?

The command derived from the model is:

sem (csp2012 -> csp2013) (csp2012 -> csrrep2012) (csp2012 -> csrrep2013) (csp2013 -> csp2014) (csp2013 -> csrrep2013) (csp2013 -> csrrep2014) (csp2014 ->
> csrrep2014) (csrrep2012 -> csrrep2013) (csrrep2012 -> roa2012) (csrrep2012 -> roa2013) (csrrep2013 -> csrrep2014) (csrrep2013 -> roa2013) (csrrep2013 ->
> roa2014) (csrrep2014 -> roa2014) (roa2012 -> roa2013) (roa2013 -> roa2014), nocapslatent

[Quint Koevoets]

I have generated the model by building it with SEM. I have added the model in the attachment. I would like to ask where I can fit in my covariate variable rdexpenses? Since it is not an extra wave I assume it can be put under y1 y2 y3 and make paths for example like: rd1 -> y1 rd1>y2 rd1>rd2 etc.? And are the covariates latent or observed variables?

The command derived from the model is:

sem (csp2012 -> csp2013) (csp2012 -> csrrep2012) (csp2012 -> csrrep2013) (csp2013 -> csp2014) (csp2013 -> csrrep2013) (csp2013 -> csrrep2014) (csp2014 ->
> csrrep2014) (csrrep2012 -> csrrep2013) (csrrep2012 -> roa2012) (csrrep2012 -> roa2013) (csrrep2013 -> csrrep2014) (csrrep2013 -> roa2013) (csrrep2013 ->
> roa2014) (csrrep2014 -> roa2014) (roa2012 -> roa2013) (roa2013 -> roa2014), nocapslatent

[wbuchanan]

Quint Koevoets at this point, the leaders in mediation modeling would likely (and without much dispute) be David MacKinnon and Andrew Hayes. I can't tell you where to place your covariates, because it depends on your theoretical rationale and reasoning. For example, you could specify a covariate relationship with the dependent variable only, the covariate may indirectly affect the dependent variable through the mediator, the covariate may affect the independent variable (which would then itself be a mediating variable), or it could be some combination of effects. Think of the path diagram like a process model. You are trying to use directed relationships among variables that best represents your hypothesized understanding of the data generating process.

With regards to the question sent outside of the forum, I have no way of knowing whether or not your variables are observed/latent. Can you directly measure the variables that you are using? If so, they are more than likely observed. If not (e.g., depression, ebility, etc...) they are by definition latent - or unobservable. If you have latent variables, you should be specifying the respective measurement models in the structural equation model itself.

[Quint Koevoets]

wbuchanan thank you for your reply and clear insight. I have hypothesized that the covariates are affecting the dependent variable (since high rdexpenses (covariate) result in higher firm performance, thus ROA (Y). Therefore I think it will be correct when placing the covariate per year (x1, x2 and x3) beneath the Y row (in the visual representation of the model) and establish paths with y1 y2 and y3. Please do not hesitate to comment on this if it seems incorrect or whatsoever.

My variables are all able to be directly measured so therefore I have used these as observed variables.

Thanks in advance!

[wbuchanan]

Quint Koevoets I'm trying to infer the meaning from the variable names, so this may not be completely correct. Are you suggesting that higher research and development expenses directly affect a firms performance? Or, does the investment in R&D affect the firms performance indirectly (e.g., Apple spends tons on R&D but I think one could argue that the ROI from this investment affects their performance indirectly through the introduction of new products being brought to market that do not exist)? Mediation is all about an indirect effect of X on Y (in the simplest form) through some intermediate variable (M). However, not all of the related variables would need to be mediated. So in the example below, the path model is suggesting the the number of competing firms has an indirect effect on a firm's share price through existing market share (e.g., fewer competing firms should reduce supply or at least allow fewer firms to control the supply curve), operational expenses (e.g., as the number of firms grows there is less human capital available for staffing positions and likely a higher labor cost due to the increase in competition), and R&D (e.g., as there are more firms competing in the same space the firm would need to invest more heavily to stay ahead of the competition). Existing market share would only have a direct effect, and operational expenses has both a direct and indirect (e.g., mediated by R&D) effects on share price. Granted this is purely for illustrative purposes, but should help you to think through how the covariates affect the outcome of interest.

[Quint Koevoets]

By the way, the function estat teffects calculates the total, direct and indirect effects and therefore indicate whether M mediates X>Y?

[wbuchanan]

Quint Koevoets all any statistical software can do is the math. The interpretation and inferences are completely a human thing. You need to use your understanding of the underlying math/models in conjunction with the theory to interpret the results derived from fitting the model to the data.

[Quint Koevoets]

wbuchanan I had not seen your second post about the covariate when I asked about the estat effects, my apologies and thank you for your comment.

I am investigating the indirect effect of Corporate Social Performance (CSP) on firm performance through CSR reputation. I have found in other studies that it is essential to control for R&D investments when investigating the CSP - performance link. Since the R&D intensity of a firm significantly affect the firm performance of a firm. Therefore, to generate more reliable results when investigating a relationship with CSP and firm performance it is advised to control for the amount of R&D investments a firm has done in that year. Therefore, I thought it was a good idea to control for the R&D expenses like this in my model:

[Quint Koevoets]

wbuchanan okay thank you for your reply. I meant that in the book of Mckinnon it is explained how to calculate the paths and therefore the indirect effect etc. But this math can also be done by stata by using eftat teffects right?

I hope I am able to correctly interpret the results for the second autogegressive model of Mckinnon (2008).

[Quint Koevoets]

By the way, a example in the paper of Mcwilliams and Siegel (2000) is that with their model CSP is positively significantly affecting firm performance, however this relationship is neutral when R&D intensity is controlled in the model.

Therefore, I thought it was an good idea to control for R&D intensity of firms on its firm performance.

[Alan Riley (StataCorp)]

A quick administrative note on this thread -- the forum software incorrectly flagged some of the posts by Quint Koevoets in it as spam, resulting in duplicate copies of those posts. I apologize for the inconvenience.