# All pages

# Estimation of the unstandardized effects of causal models

In many cases, researchers are interested in the unstandardized coefficients, which can be obtained by analyzing the raw data or the covariance matrix. In this chapter, we will therefore show how the analysis can be done, starting with the covariance matrix of the observed variables of interest. The first step is the estimation of the covariance matrix weighted for design weights, as we did for the correlation matrix. The second step is correction of the covariances and variances for measurement error. After that, the unstandardized coefficients can be estimated as before.

The covariance matrix can easily be estimated using any statistical program. For those interested in reproducing the results of this module, the procedure and the syntax are provided in the following links for both SPSS and Stata, using a dataset especially prepared for this module.

- In the following link, you will find the dataset ‘CME data – ESSround 6’.
Open this dataset in SPSS:1GET FILE='C:\...\CME data_ESSround6.sav'.
- First, select the cases under study in our analysis. They concern the whole British population. Therefore, from Data in SPSS, select ‘Select Cases…’. To limit the analysis to Great Britain, choose ‘If condition is satisfied, select the variable ‘Country’ and insert the following notation: cntry = ‘GB’.
COMPUTE filter_$=(cntry="GB").VARIABLE LABELS filter_$ 'cntry="GB" (FILTER)'.VALUE LABELS filter_$ 0 'Not Selected' 1 'Selected'.FORMATS filter_$ (f1.0).FILTER BY filter_$.EXECUTE.
- Under Data, you have to weight the cases using design weights. Select ‘Weight cases’ and weight the cases by the variable ‘Design weight [dweight]’.
WEIGHT BY dweight.
- To obtain the covariance matrix, choose ‘Correlate’ from Analyze and then click ‘Bivariate…’. From the list, select the variables in the following order: Satdem [stfdem], Free [fairelecc], Critic [oppcrgvc], Equal [cttresac], LRplace [lrscale] and Inc [hinctnta]. Once the variables are selected, in Options choose ‘Cross-product deviations and covariances’ to obtain the statistics of those variables and also choose the option ‘Exclude cases listwise’ to obtain the results for the same cases in the sample.
CORRELATIONS/VARIABLES=stfdem fairelcc oppcrgvc cttresac lrscale hinctnta/PRINT=TWOTAIL NOSIG/STATISTICS XPROD/MISSING=LISTWISE.

This procedure should lead to the result in the following table:

- In the following link, you will find the dataset ‘CME data – ESSround 6’
Open this dataset in Stata:2use "C:\...\CME data_ESSround6.dta", clear
- Select the cases under study. They concern the whole British population. Therefore, in Stata we can use the command ‘keep if’ and indicate that we will keep all observations that for the variable ‘Country (cntry)’ have the value ‘GB’.
keep if cntry=="GB"
- Performing some tabulations of our variables of interest before seeing the descriptives, we see that the variables Satdem [stfdem], Free [fairelecc], Critic [oppcrgvc], Equal [cttresac], LRplace [lrscale] and Inc [hinctnta] have Refusal and Don’t know values, which should be assigned to system missing. This can be done using the command ‘mvdecode’.
mvdecode stfdem fairelcc oppcrgvc cttresac lrscale hinctnta, mv(99)mvdecode stfdem fairelcc oppcrgvc cttresac lrscale hinctnta, mv(88)mvdecode stfdem fairelcc oppcrgvc cttresac lrscale hinctnta, mv(77)
- To obtain the descriptive statistics in Stata, we have used the command ‘corr’. Using this command, select the seven variables under analysis in the following order: Satdem [stfdem], Free [fairelecc], Critic [oppcrgvc], Equal [cttresac], LRplace [lrscale] and Inc [hinctnta]. Here, the design weights have been applied using the command ‘aweight’. Furthermore, in order to obtain the covariance matrix we add the notation ‘cov’.
corr stfdem fairelcc oppcrgvc cttresac lrscale hinctnta [aweight=dweight], cov

This procedure should lead to the result in the following table:

^{3}for Great Britain corrected by design weights (n=1468)

^{4}.

- [1] The following illustration and results are based on the SPSS 19 software version: IBM Corp. Released 2010. IBM SPSS Statistics for Windows, Version 19.0. Armonk, NY: IBM Corp.
- [2] The following illustration and results are based on the Stata 12 software version: StataCorp. 2011.
**Stata Statistical Software: Release 12**. College Station, TX: StataCorp LP. - [3] ESS Round 6: European Social Survey Round 6 Data (2012). Data file edition 2.0. Norwegian Social Science Data Services, Norway – Data Archive and distributor of ESS data.
- [4] SPSS adjusts the sample size on the basis of the design weights. Their adjusted sample size is 1424. However, for our illustration, we will stick to the original sample of 1468, which is the actual number of people that answered the questions.

# Correction for measurement errors in the covariance matrix

The next step is the computation of the correction of the covariances for common method variance. As before, this will be done for the variables sharing the same method. The coefficients that should be corrected for common method variance are highlighted in the following table.

The starting point for the computations of these corrections is the cmv of the standardized variables computed above in Table 4.4 and reproduced here in Table 6.3.

To obtain the common method variance for the same unstandardized variables, these cmv have to be multiplied by the standard deviations for the specific variables.

Unstandardized cmv = (Standardized cmv) s_{i}s_{j} | equation 6.1 |

From the data, we get:

_{Satdem}= 2.152, s

_{Free}= 1.879, s

_{Critic}= 1.946, s

_{Equal}= 2.768, s

_{LRplace}= 1.813, s

_{Inc}= 2.986

The result of this computation is the common method variance of the unstandardized variables, which has to be subtracted from the estimated covariances for the same variables to obtain the corrected covariances. The different steps are presented in Table 6.4.

These results are introduced in the proper place in the covariance matrix. The resulting matrix is presented in Table 6.5.

The last step is to correct the variances for all variables for measurement errors. This is done by substituting the originally estimated variances of all variables by the variances corrected for measurements errors. The variance coefficients that should be corrected in this case are highlighted in the following table.

Because 1-q^{2} is the proportion of error variance in the observed variables, the variances corrected for errors can be computed by multiplying the variances by the quality (from Table 3.3) of the variables. The result of these steps is presented in Table 6.7.

The results are introduced in the diagonal in the covariance matrix. The result is presented in Table 6.8.

Now that the covariance matrix has been corrected, the analysis with correction for measurement errors can start.

# Estimation of the unstandardized coefficients of the causal model

Given that Table 6.8 contains the covariances and variances corrected for measurement errors, the next step is now the same as that done in Chapter 5. In fact, the inputs for the programs are exactly the same except for the data we use as the starting point.

Below, we will illustrate how to run the estimation of the unstandardized coefficients of the causal model specified above (Figure 6.1) using both LISREL1 and Stata2. As both programs provide very similar results, please select which program you want to continue the analysis with.

In Figure 6.2, all effects have been indicated using the symbols from LISREL. The betas (be) represent the effects of the explanatory (endogenous) variables (i.e. Free, Critic and Equal) on satisfaction with democracy (i.e. Satdem). For example, be(1,2) indicates the effect of the variable freedom and fairness of elections on satisfaction with democracy. Similarly, the gammas (ga) represent the effects of the control (exogenous) variables (i.e. LRplace and Inc) on the other variables in the model. For example, ga(1,1) indicates the effect of the control variable Left-right placement on satisfaction with democracy, while ga(3,2) indicates the effect of the other control variable, Income (i.e. Inc), on the variable equality by law (i.e. Equal). The effect of the variable Inc on Satdem is specified by a dashed line because it represents an effect that has been omitted because it was not significant in the analysis with correction for measurement errors.

We do not expect the control variables to completely explain the correlations that exist between the evaluation of democracy questions. We can therefore expect correlations between the disturbance terms (ζ_{2}-ζ_{4}) of these variables. These correlations are not indicated in the model, but are denoted in LISREL by ps(3,2), ps(4,2) and ps(4,3), while the variances of the disturbances are denoted as ps(1,1), ps(2,2), ps(3,3) and ps(4,4). For details of the procedure, we refer to the LISREL manual [Jör96] and introductions to the program LISREL [Sar84]. First, the LISREL input for this analysis without corrections is presented in Syntax 6.1. Next, we present the same input corrected for measurement errors (see Syntax 6.2).

**Syntax 6.1:**The LISREL syntax for the estimation of the unstandardized coefficients of the causal model without correction for measurement errors*

*Note that the effect of Income on Satdem (ga(1,2)) has not been introduced in this syntax without correction for measurement errors.

**Syntax 6.2:**The LISREL syntax for the estimation of the unstandardized coefficients of the causal model with correction for measurement errors

The most important point is that the coefficients that have to be estimated are presented in the lines starting with ‘free’. Comparing these two inputs, we see that only the matrix with the data to be analysed has been changed. Focusing on the input for the model with correction for measurement errors, the effects will be estimated on the basis of the corrected covariance matrix in Table 6.8 (i.e. the matrix with the correction for measurement errors).

The nice feature of this approach, correcting the correlations for measurement errors before estimating the effects, is that the input for the analysis is exactly the same with and without correction for measurement errors, except for the matrix of variances and covariances that is used in the analysis. This point is illustrated in the input for the analyses with and without correction for measurement errors presented in Syntaxes 6.1 and 6.2.

However, analysing the matrix without correction for measurement errors, the program indicates that the fit of the model is not good. This suggests that the effect, ga(1,2), of the control variable Income on satisfaction with democracy has to be introduced in the model. If we do so, this model also fits well to the data and we obtain the unstandardized results presented in Table 6.9.

In Syntaxes 6.1 and 6.2, we can observe that all effects have been indicated using the Stata notation [Aco13]. We can expect variances between the disturbance terms of the evaluation of democracy questions, which are denoted in Stata as e.free*e.critic, e.free*e.equal and e.critic*e.equal. Comparing the two inputs, we see that only the matrix with the data to be analysed has been changed. Focusing on the input for the model with correction for measurement errors, the effects will be estimated on the basis of the corrected covariance matrix in Table 6.8 (i.e. the matrix corrected for measurement errors).

**Syntax 6.1:**The Stata syntax for the estimation of the unstandardized coefficients of the causal model without correction for measurement errors**

**Note that the effect of Income on Satdem has not been introduced in this syntax without correction for measurement errors.

**Syntax 6.2:**The Stata syntax for estimation of the unstandardized coefficients of the causal model with correction for measurement errors

The nice feature of this approach, correcting the correlations for measurement errors before estimating the effects, is that the input for the analysis is exactly the same with and without correction for measurement errors, except for the matrix of variances and covariances that is used in the analysis. This point is illustrated in the input for the analyses with and without correction for measurement errors presented in Syntaxes 6.1 and 6.2.

However, analysing the matrix without correction for measurement errors, the program indicates that the fit of the model is not good. This suggests that the effect of the control variable income (Inc) on satisfaction with democracy has to be introduced in the model. If we do so, this model also fits well to the data and we obtain the estimates of the unstandardized coefficients presented in Table 6.10.

If we compare the unstandardized results with and without correction for measurement error, we see first of all that the model is different. After correction for measurement errors, the effects of the control variable (Inc) on satisfaction with democracy is not significantly different from zero, while, without correction for measurement errors, this effect is needed to achieve a good fit of the model. In the latter case, we say that this variable has a direct effect on satisfaction with democracy, while in the former analysis we have to conclude that there is no direct effect, only an indirect effect.

Furthermore, comparing Figures 6.3 and 6.4, we see that nearly all other unstandardized effects after correction for measurement error are bigger than without correction for errors. In this case, the difference is in fact a little bit smaller than it was for the standardized solution, but still substantively rather important. All significant effects are indicated by an asterisk (*) in the figures. Using this approach, we see that, in both analyses, the effect of left-right placement on freedom to criticise is not significant on the 5% significance level. The explained variance in Satdem increases from 22.7% to 46.6%3, as before.

This example illustrates again how different the results can be if one does or does not correct for measurement errors, even if the covariance matrix is used. It also shows that the analysis with correction for measurement errors is not more difficult than the analysis without correction for errors. It is only necessary to correct the covariances before starting the analysis.

## Exercise 6.1

Compute the corrected covariance matrix for the variables introduced in exercise 3.1. The covariance matrix without corrections is provided in the tables below, together with the quality predictions obtained in exercise 3.1 using SQP, the standardized cmv obtained in exercise 4.3 and the descriptive statistics of the variables.

Firstly, we have to compute the unstandardized cmv for the variables measured by a common method in order to correct the covariances between these variables. This applies to the variables Economy, Culture and Better, which use 11-point, item-specific scales. The computation of the standardized cmv was already done in exercise 4.3. With these results, we are able to compute the unstandardized cmv based on the following formula:

_{i}s

_{j}

_{B40,B38}= 0.094 * 1.817 * 1.908 = 0.327

_{B40,B39}= 0.110 * 1.817 * 1.942 = 0.388

_{B38,B39}= 0.117 * 1.908 * 1.942 = 0.433

Secondly, we have to subtract the unstandardized cmv from the covariances corrected for common method variance:

_{B40,B38}= 1.850 – 0.327 =

**1.523**

_{B40,B39}= 1.966 – 0.388 =

**1.578**

_{B38,B39}= 1.965 – 0.433 =

**1.532**

Thirdly, correct the variances on the diagonal by removing the error variance or putting the product of the quality obtained from SQP on the diagonal:

_{B37}= 0.721 * 0.763 =

**0.550**

_{B40}= 3.300 * 0.639 =

**2.109**

_{B38}= 3.641 * 0.702 =

**2.556**

_{B39}= 3.770 * 0.641 =

**2.417**

By imputing the corrected variances on the diagonal from the table above, the covariance matrix corrected for measurement errors has been obtained.

## Exercise 6.2

Repeat the estimation of the causal model for the same unstandardized variables. The covariance matrix with corrections obtained in exercise 6.1 is presented in the table below. Furthermore, the covariance matrix without corrections was provided in the previous exercise. Use this information to compute, either in LISREL or Stata, the results for the analysis of the explanation of the opinion about immigration by people from outside Europe to the Netherlands with and without correction for measurement errors presented in the next figure.

The model to be estimated is:

**LISREL syntax for the estimation of the unstandardized causal model without corrections for measurement errors:**Unstandardized causal analysis without correctionsdata ni=4 no=1801 ma=cmcm0.721-0.542 3.300-0.652 1.850 3.641-0.561 1.966 1.965 3.770labelsimpcntr imwbcnt imbgeco imuecltmodel ny=2 nx=2 be=fu,fi ga=fu,fi ps=sy,fifree be(1,2)free ga(1,1) ga(1,2) ga(2,1) ga(2,2)free ps(1,1) ps(2,2)pdout nd=3 sc**LISREL syntax for the estimation of the unstandardized causal model with corrections for measurement errors:**Unstandardized causal analysis with correctionsdata ni=4 no=1801 ma=cmcm0.550-0.542 2.109-0.652 1.523 2.556-0.561 1.578 1.532 2.417labelsimpcntr imwbcnt imbgeco imuecltmodel ny=2 nx=2 be=fu,fi ga=fu,fi ps=sy,fifree be(1,2)free ga(1,1) ga(1,2) ga(2,1) ga(2,2)free ps(1,1) ps(2,2)pdout nd=3 sc

**Stata syntax for the estimation of the unstandardized causal model without corrections for measurement errors:***Unstandardized causal model without correction for measurement errorsclear allssd init impcntr imwbcnt imbgeco imuecltssd set observations 1801*Covariance matrix#delimit ;ssd set covariances0.721\-0.542 3.300\-0.652 1.850 3.641\-0.561 1.966 1.965 3.770;#delimit crsave ssdmatrix.dat, replace*Causal modelclearuse ssdmatrix.datssd listsem (impcntr <- imwbcnt imbgeco imueclt) ///(imbgeco imueclt -> imwbcnt)estat eqgof**Stata syntax for the estimation of the unstandardized causal model with corrections for measurement errors:***Unstandardized causal model with correction for measurement errorsclear allssd init impcntr imwbcnt imbgeco imuecltssd set observations 1801*Covariance matrix#delimit ;ssd set covariances0.550\-0.542 2.109\-0.652 1.523 2.556\-0.561 1.578 1.532 2.417;#delimit crsave ssdmatrix.dat, replace*Causal modelclearuse ssdmatrix.datssd listsem (impcntr <- imwbcnt imbgeco imueclt) ///(imbgeco imueclt -> imwbcnt)estat eqgof

The figure at the top presents the results for the unstandardized causal model without corrections. Again, we see from this model that all effects are significant. However, compared with the standardized results obtained in exercise 5.2, we can see that the effects are now smaller. The second figure shows the results for the unstandardized causal model with corrections. Comparing these two models, we see that the increase in the effects is small. This was explained in the last chapter. The variance explained with and without corrections for the unstandardized causal model is equal to the variance explained in the standardized causal model.

- [1] The following illustration and results are based on the LISREL 8.7 software version: Jöreskog, K.G. & Sörbom, D. (2004). LISREL 8.7 for Windows [Computer software]. Skokie, IL: Scientific Software International, Inc.
- [2] The following illustration and results are based on the Stata 12 software version: StataCorp. 2011.
**Stata Statistical Software: Release 12**. College Station, TX: StataCorp LP. - [3] The explained variance can be obtained in LISREL from the section
**Squared Multiple Correlations for Structural Equations**(R^{2}). Similarly, in Stata the command**estat eqgof**will show you the value of R^{2}.

- [Aco13] Acock, A. C. (2013). Discovering Structural Equation Modeling Using Stata, Revised Edition.
*Stata press*. - [Jör96] Jöreskog, K. G. and Sörbom, D. (1996). LISREL 8 User’s Reference Guide.
*Scientific Software International*. - [Sar84] Saris, W. E. and Stronkhorst, L. H. (1984). Causal modelling in nonexperimental research: an introduction to the LISREL approach.
*Sociometric Research Foundation*.