**3.1 Psychometric properties of the instrument**

#### **Reliability**

The reliability is calculated by the internal consistency of scales. When calculating the internal consistency reliability of the four instruments, the Cronbach α is 793, which represents an acceptable reliability. When calculating the reliability for each of the four instruments, the results are variable, being generally adequate except for PRAESPE with a Cronbach α of .581, which is rather low, with regard to the rest of the scales α Cronbach; for the PRAESPO = 861; for the PRAESPA = 788 and, for the PRAESAE = 843.

#### **Validity**

To study the validity of the PRAES instrument, two types of analysis were carried out: content or apparent validity and construct validity.

#### **Content validity of PRAES**

From the revision of the international theoretical and empirical studies published in recent years, we can reasonably assert that the apparent or content validity of the PRAES is acceptable. In this review, we summarize the findings, the empirical evidence and the theoretical conceptualizations, in the four major components that relate to the role of the teachers' practice in teaching writing.

The first component, the PRAESPO is concerned with the way teachers relate to the multiple components of teaching writing, such as motivation, planning and review, family, and the teacher's practice and training. This type of instrument, based on obtaining the opinion of teachers, has been used previously, in part, by Graham et al., (2001).

The second component, PRAESPE, refers to the theoretical approach followed by teachers at the schools involved in the study, which can be dichotomized into two major clusters along a continuum. At one extreme is formal education and at the other we find natural learning. It is understood that the different teachers are located along the continuum, with some elements of both approaches. This type of component has been studied previously in several studies by Graham et al., (2001).

Moreover, the component related to the specific practice undertaken by the teacher in the classroom in the teaching of writing, PRAESPA has to do with elements that include the deployment of skills, the use of specific procedures, the performance of the various activities, the use the appropriate materials or the use of diverse textual genres. Some of these aspects have been measured and studied by Graham et al., (2001).

Finally, we look at the self-perception component of efficacy which the teacher holds regarding their performance as a writing instructor/teacher, the PRAESAE. This component refers to the self-perception of the teacher, is related by Bandura (2005), and we followed his guidelines in the construction of that component.

#### **Construct validity of PRAES**

202 Learning Disabilities

dates and times for teachers to complete the protocols. The collection of the sample was performed directly by the two researchers in order to ensure the reliability and validity of the instruments in the collection of information. Data collection was conducted over a period of five months. It is necessary to highlight the effort exerted both in the use of persuasion and communication techniques to synchronize more effectively with the agents under study and to develop their awareness regarding research, as well as the physical

Once the field work was completed and all the protocols were collected, a total of 137, the codification and computerization of the data in an Excel data matrix were carried out. This matrix was transformed into a SPSS version 13.0 matrix which is available in the ULE for conducting statistical analysis. Then the preparation of tables, graphs and other data used for the presentation of results and to provide empirical evidence was carried out, including the interpretation of the data and the identification and extraction of conclusions.

The analysis and findings set out below focus on four specific points, (i) an analysis of the psychometric properties of PRAES, including its internal consistency reliability and validity of both the content and construct, (ii) an analysis of the descriptive data which highlight the averages of the measures and variables, (iii) a linear regression analysis (linear regression

The reliability is calculated by the internal consistency of scales. When calculating the internal consistency reliability of the four instruments, the Cronbach α is 793, which represents an acceptable reliability. When calculating the reliability for each of the four instruments, the results are variable, being generally adequate except for PRAESPE with a Cronbach α of .581, which is rather low, with regard to the rest of the scales α Cronbach; for

To study the validity of the PRAES instrument, two types of analysis were carried out:

From the revision of the international theoretical and empirical studies published in recent years, we can reasonably assert that the apparent or content validity of the PRAES is acceptable. In this review, we summarize the findings, the empirical evidence and the theoretical conceptualizations, in the four major components that relate to the role of the

The first component, the PRAESPO is concerned with the way teachers relate to the multiple components of teaching writing, such as motivation, planning and review, family, and the

Furthermore, the limitations and prospects for future study were determined.

models), and (iv) a multivariate analysis of variance (general linear models).

the PRAESPO = 861; for the PRAESPA = 788 and, for the PRAESAE = 843.

**3.1 Psychometric properties of the instrument** 

content or apparent validity and construct validity.

effort and the costs of work.

**3. Results** 

**Reliability** 

**Validity** 

**Content validity of PRAES** 

teachers' practice in teaching writing.

We obtained the estimate of the construct validity using a factor analysis of the principal components with a normalized varimax rotation.

The analysis allowed the extraction of five factors or principal components that only partially conform to the structure of the test following the validity of content which was developed from a review of international theoretical and empirical studies. In general, the factors extracted from the analysis did not coincide exactly with the four components of PRAES, although we can say that they were largely close, indicating at least partially an acceptable construct validity.

Firstly, we obtained a factor which has been labeled as general because it includes the saturation of the factorial weights of the total number of PRAES scales. The general factor explains 33.20% of the total variance predicted and is saturated by the weight factor for the PRAESPO, the subtotal of motivation (-.738) and the family subtotal (-.882), plus the PRAESPA with the texts subtotal (.765) and the activity subtotal (.813), as well as the personal self-efficacy subtotal (.569) and the general self efficacy subtotal (.479), although in the latter case, as can be seen, the totals are of a lesser magnitude.

Secondly, we obtained a factor which can be called theoretical approach which includes self-saturation of the factor weights of the total number of scales in the PRAESAE. This factor explains 22.49% of the total variance and is saturated by the factor weights corresponding to the PRAESPE, the formal education subtotal (.929) and the natural learning subtotal (-.582) with the PA PRAES with its skills sub-totals (.905) and the materials subtotals (.446), plus the personal self-efficacy totals (.657) and general selfefficacy totals (.580). These lower factorial weights contribute less to this factor's configuration. Thirdly, we obtained a factor we have called opinion, since it includes the saturation factor of the total weights of the various PRAESPO scales. This factor explains

Diagnosis of Teachers' Practice in the Teaching of Written Composition 205

98 23 67 46.92 11.20

111 1 43 23.43 11.75

137 -1 6 3.4 1.96

113 1 43 18.84 12.77

132 12 35 29.65 3.96 35

136 4 28 24 3.37 28

135 10 41 32.54 4.52 42

136 4 21 14.91 2.78 21

129 6 26 18.76 3.30 28

124 13 54 39.99 7.76 54

124 19 48 31.67 6.57 54

125 14 33 24.46 4.12 35

94 7 30 16.89 4.99 42

132 9 28 19.12 4.08 28

109 2 12 6.02 2.12 77

120 6 28 16.03 5.18 28

119 35 90 62.37 9.82 90

115 24 50 36.34 5.57 54

deviation

Maximum possible score

VARIABLES COMPONENTS N Min. Max. Average Standard

Opinion, motivation subcomponent.

Opinion, practice subcomponent.

Opinion, family subcomponent.

Opinion, teacher

Natural learning approach subcomponent.

Natural learning approach subcomponent.

Classroom, skills subcomponent.

Classroom, texts subcomponent.

subcomponent.

SELF-EFFICACY Personal self-efficacy

Classroom, materials subcomponent.

Classroom, procedure subcomponent.

Classroom, activities subcomponent.

General self-efficacy subcomponent.

Table 3. Description of the statistics of the general variables and instruments

training subcomponent.

Opinion, planning and review subcomponent.

AGE OF THE TEACHER

SCHOOL YEAR TAUGHT

YEARS OF TEACHING

YEARS TEACHING SPANISH LANGUAGE

OPINION

APPROACH

CLASSROOM

16.22% of total variance and is saturated by the factor weights for the PRAESPO, the practice subtotal (.875) and the teacher training subtotal (.896), plus the PRAESPE natural learning subtotal (.640).

A further factor, called activities was obtained which includes the saturation factor of the total weights of the various PRAESPA scales. This factor explains 8.80% of total variance and is saturated by the factor weights corresponding to the PRAESPA, the procedures subtotal (.978) and the materials subtotal (.649), plus the PRAESPO motivation subtotal (.442). The latter with less weight and therefore make a smaller-scale contribution this factor. In fifth place, there is the review-planning factor, so called because it includes saturation of the weights of the total factors in one of the PRAESPO scales. This factor explains 7.23% of total variance and is saturated by the factor weights of the PRAESPO, the planning and review subtotal (.939), plus the PRAESAE general self-efficacy subtotal (-.408).

## **Descriptive data**

In this section we give an overview of the descriptive results. We present the averages of the measures and variables in Table 3. This table contains data on the number of participants, the minimum and maximum scores obtained for each variable and the respective averages and standard deviations, as well as the highest scores for each of the variables.

As shown in Table 3 and Figure 1, the data generally showed a positive result for each of the variables included in the instruments. For example, in the motivation sub-component of PRAESPO the average score was 29.65, which exceeded the median potential in this variable, and a maximum score of 35 as obtained. Furthermore, the average years of teaching experience was 23.43, which indicates great teaching experience, with 18.84 of those years teaching Spanish language, which suggests the teachers were highly knowledgeable regarding instructional aspects, which is probably related to the teaching of writing.

Fig. 1. Descriptive results in PRAES (teachers´ practice)

16.22% of total variance and is saturated by the factor weights for the PRAESPO, the practice subtotal (.875) and the teacher training subtotal (.896), plus the PRAESPE natural

A further factor, called activities was obtained which includes the saturation factor of the total weights of the various PRAESPA scales. This factor explains 8.80% of total variance and is saturated by the factor weights corresponding to the PRAESPA, the procedures subtotal (.978) and the materials subtotal (.649), plus the PRAESPO motivation subtotal (.442). The latter with less weight and therefore make a smaller-scale contribution this factor. In fifth place, there is the review-planning factor, so called because it includes saturation of the weights of the total factors in one of the PRAESPO scales. This factor explains 7.23% of total variance and is saturated by the factor weights of the PRAESPO, the planning and review subtotal (.939), plus the PRAESAE general self-efficacy subtotal

In this section we give an overview of the descriptive results. We present the averages of the measures and variables in Table 3. This table contains data on the number of participants, the minimum and maximum scores obtained for each variable and the respective averages

As shown in Table 3 and Figure 1, the data generally showed a positive result for each of the variables included in the instruments. For example, in the motivation sub-component of PRAESPO the average score was 29.65, which exceeded the median potential in this variable, and a maximum score of 35 as obtained. Furthermore, the average years of teaching experience was 23.43, which indicates great teaching experience, with 18.84 of those years teaching Spanish language, which suggests the teachers were highly knowledgeable regarding instructional aspects, which is probably related to the teaching of

*Figure 1 Descriptive results in PRAES (teachers' practice)*

and standard deviations, as well as the highest scores for each of the variables.

Fig. 1. Descriptive results in PRAES (teachers´ practice)

learning subtotal (.640).

(-.408).

writing.

**Descriptive data** 


Table 3. Description of the statistics of the general variables and instruments

Diagnosis of Teachers' Practice in the Teaching of Written Composition 207

= .480), for the PRAESPA and its skills subcomponent (p = .024, n = ² 564). The same natural theoretical approach without clusters, was as expected (p = .001, n ² = .938), just as the

SIGNIFICANT VARIABLES λ F(2,18) p ŋ² 0.001 46.465 0.021 0.998

Theoretical natural approach PRAESPA: Skills Subcomponent 0.024 0.564

 Personal self-efficacy 0.024 0.565 Λ F (1.10) p ŋ²

Personal self-efficacy PRAESPA: Materials Subcomponent 0.092 0.258

Personal self-efficacy 0.002 0.626

Table 4. Results of the multivariate variance analysis of the statistically significant variables

*Natural approach (low\_average\_high)*

The tests concerning the post hoc contrasts, however, generally do show significant differences between the clusters of this variable. In terms of personal self-efficacy, the test for the inter-subject effects indicates statistical significance for PRAESPO variables, with the

*Formal approach Skills Natural approach Personal selfefficacy*

*Serie1 Serie2 Serie3*

of the PRAES in three conglomerates

Fig. 2. Natural approach (low\_average\_high)

Theoretical formal approach 0.053 0.480

Theoretical natural approach 0.001 0.938

PRAESPO: Family Subcomponent 0.038 0.365

PRAESPA: Texts Subcomponent 0.001 0.779

0.001 1169.051 0.023 1

personal self-efficacy variable (p = .024, n ² = .565), as shown in Figure 2.

#### **Multiple linear regression analysis**

The aim is to predict the dependent variable (predicted variable) from the rest of the predictor variables. The total of the four scales are taken as predictors and each of the predicted variables is taken in turn, including the identification of other variables, such as the way in which the teacher teaches. When the educational year taught is taken as a dependent variable or as a variable predicted from the PRAES set of measures and their subscales in the hierarchical multiple regression analysis model with a single statistically significant regression coefficient (adjusted R ² = .313) is obtained, step by step. One single variable was included in the model, from the PRAESPA – the procedure sub-component [β = .613, t=-2452, p = .034]. The rest of the variables were excluded from the model, as they did not achieve statistical significance. Moreover, when attempting to predict the teacher's formal theoretical approach, an adjusted R ² of .616 was obtained, which is very high although it only generates one model or regression equation with the predictor variable PRAESPA in the skills subcomponent [β = 807 t = 4315, p = .002]. There were no other significant variables, which meant they were consequently excluded from the model. Similarly, the theoretical prediction of natural teaching can be realized with two regression equations. The first model gives an adjusted R ² of .331, with a significant variable found in personal self-efficacy [β = - .626, t=-2540, p = .029], and in the second model, an adjusted R ². 556 was obtained, the texts subcomponent being significant in addition to the personal selfefficacy subcomponent of the PRAESPA [β = .774, t= 2465, p = .036]. As a result, we excluded the remaining variables in the step by step multiple regression analysis. In summary, regression analysis provides statistically significant data for predicting each of the PRAES variables from the other variables, which is very interesting as regards its predictive validity. Furthermore, it indicates the great predictive potential of PRAES, as it allows us to obtain knowledge about variables that are not always of the same in nature as the rest. For example, the PRAES includes an opinion subcomponent concerning selfefficacy another that outlines the type of behaviors or activities performed by teachers in the classroom and a third concerning the theoretical approach followed by teachers in writing instruction. This different character, which provides such high regression coefficients, is a good indicator, which supports the integrity of the PRAES.

#### **Multivariate analysis of variance (general linear models)**

It is important to highlight that the analysis to distinguish the role or influence of the different grouping variables (teachers' gender, years of teaching experience, education centre, and the school year taught), does show statistical significance in the multivariate contrasts. Each of the PRAES variables used as an inter-subject factor were not statistically significant in the multivariate contrasts, except in the case of the teacher's natural theoretical approach. This variable, when grouped into three clusters, with a multivariate variance contrast is statistically significant and also shows a large effect size [λ = 001 F (2.18) =46,465, p = .021, n ² = 998]. This is similar to the personal self-efficacy variable when it is grouped into three clusters and shows a large effect size [λ = 001, F (1, 10) = 1169.051, p = .023, n ² = 1]. The tests of the inter-subject effects of personal self-efficacy and natural theoretical approach indicate there are some significant variables (see Table 4).

As regards the natural theoretical teaching approach, the evidence of the inter-subject effects indicate statistical significance for variables of the formal theoretical approach (p = .053, n ²

The aim is to predict the dependent variable (predicted variable) from the rest of the predictor variables. The total of the four scales are taken as predictors and each of the predicted variables is taken in turn, including the identification of other variables, such as the way in which the teacher teaches. When the educational year taught is taken as a dependent variable or as a variable predicted from the PRAES set of measures and their subscales in the hierarchical multiple regression analysis model with a single statistically significant regression coefficient (adjusted R ² = .313) is obtained, step by step. One single variable was included in the model, from the PRAESPA – the procedure sub-component [β = .613, t=-2452, p = .034]. The rest of the variables were excluded from the model, as they did not achieve statistical significance. Moreover, when attempting to predict the teacher's formal theoretical approach, an adjusted R ² of .616 was obtained, which is very high although it only generates one model or regression equation with the predictor variable PRAESPA in the skills subcomponent [β = 807 t = 4315, p = .002]. There were no other significant variables, which meant they were consequently excluded from the model. Similarly, the theoretical prediction of natural teaching can be realized with two regression equations. The first model gives an adjusted R ² of .331, with a significant variable found in personal self-efficacy [β = - .626, t=-2540, p = .029], and in the second model, an adjusted R ². 556 was obtained, the texts subcomponent being significant in addition to the personal selfefficacy subcomponent of the PRAESPA [β = .774, t= 2465, p = .036]. As a result, we excluded the remaining variables in the step by step multiple regression analysis. In summary, regression analysis provides statistically significant data for predicting each of the PRAES variables from the other variables, which is very interesting as regards its predictive validity. Furthermore, it indicates the great predictive potential of PRAES, as it allows us to obtain knowledge about variables that are not always of the same in nature as the rest. For example, the PRAES includes an opinion subcomponent concerning selfefficacy another that outlines the type of behaviors or activities performed by teachers in the classroom and a third concerning the theoretical approach followed by teachers in writing instruction. This different character, which provides such high regression coefficients, is a

It is important to highlight that the analysis to distinguish the role or influence of the different grouping variables (teachers' gender, years of teaching experience, education centre, and the school year taught), does show statistical significance in the multivariate contrasts. Each of the PRAES variables used as an inter-subject factor were not statistically significant in the multivariate contrasts, except in the case of the teacher's natural theoretical approach. This variable, when grouped into three clusters, with a multivariate variance contrast is statistically significant and also shows a large effect size [λ = 001 F (2.18) =46,465, p = .021, n ² = 998]. This is similar to the personal self-efficacy variable when it is grouped into three clusters and shows a large effect size [λ = 001, F (1, 10) = 1169.051, p = .023, n ² = 1]. The tests of the inter-subject effects of personal self-efficacy and natural theoretical

As regards the natural theoretical teaching approach, the evidence of the inter-subject effects indicate statistical significance for variables of the formal theoretical approach (p = .053, n ²

**Multiple linear regression analysis** 

good indicator, which supports the integrity of the PRAES. **Multivariate analysis of variance (general linear models)** 

approach indicate there are some significant variables (see Table 4).

= .480), for the PRAESPA and its skills subcomponent (p = .024, n = ² 564). The same natural theoretical approach without clusters, was as expected (p = .001, n ² = .938), just as the personal self-efficacy variable (p = .024, n ² = .565), as shown in Figure 2.


Table 4. Results of the multivariate variance analysis of the statistically significant variables of the PRAES in three conglomerates

Fig. 2. Natural approach (low\_average\_high)

The tests concerning the post hoc contrasts, however, generally do show significant differences between the clusters of this variable. In terms of personal self-efficacy, the test for the inter-subject effects indicates statistical significance for PRAESPO variables, with the

Diagnosis of Teachers' Practice in the Teaching of Written Composition 209

The fact that the teachers state that they do the same, believe the same, maintain the same theoretical framework for teaching writing, with the same feelings of ability, regardless of the students' level may have several causes. It may be simply an observation of what is actually expected, which contradicts our hypothesis, and that teachers apply these patterns without considering the characteristics of students. This may be because their training is not specific or because practice in writing instruction is scarce, except in the initial levels, where the mechanical aspects are predominant, which all indicates a lack of self-regulation their practice. But it is possible that, by using such general measures rather than more direct measures of the teachers' actions, which consequently reflects the teachers' social desirability, including their own ideology, beliefs and theories, their views, all of which

If the actual teaching of writing to different levels was addressed it may produce might greater differences as it is not possible to teach students of different educational levels in the same way. In this sense, the PRAES would measure the beliefs and wishes more generally than the role of the actual practice. This instrument, which was applied at the margins of the teachers' direct activity, thus reflects components of "trait" rather than "state". What would have happened if teachers had been directly observed in the classroom in their writing instruction with their students from different educational levels? It is expected that they would have demonstrated different strategies, but this possibility must be explored in another study. As for the sample, as well as being representative, relevant and of a broadspectrum (as compared with the samples from other empirical studies), allows for the description of the kind of practice carried out by these teachers, and the gaining of valuable

Concerning the instrument applied, given the review of theoretical and empirical studies published in recent years, one can reasonably assert that the PRAES presents adequate validity and acceptable reliability. However, the questionnaire is the instrument most used

Furthermore, we can ensure that the construction of PRAES demonstrated the selective and representative collection of elements that refer to the opinion or approach components, on the specific behavior carried out by the teachers in the classroom in the teaching of writing composition and the self-efficacy component, as indicated above. It should be emphasized that no published study is known to have researched the four components together. This

Regarding the statistical analysis and as a contribution to this research study, five factors or components were obtained that, although only partially conform to the structure of the test, do have some overlap with the four components of the PRAES. Firstly, there is an overlap between the general factor, in which the teachers' both personal and general self-efficacy interact with motivation, the family and the classroom activities in the teaching of writing. Secondly, is the factor of self-efficacy and theoretical approach, since the following are interrelated, the natural and formal theoretical approach of the teacher interacts, the skills they develop and the materials used in teaching writing, identifying their personal and general self-efficacy. In addition, the third factor, opinion, interrelates the nature of the theoretical approach to teacher training and the practice developed in the teaching of

data, although these are, as previously indicated, partly limited.

in research conducted in this area.

justifies and affords relevance to this research.

were very generic.

family subcomponent (p = .038, n ² = .365). For the PRAESPA materials subcomponent (p = .092, n = ². 258), although in this case, the texts subcomponent also comes close to statistical significance (p = .001, n ² = .779), and as does personal self-efficacy (p = .002, n²=.626). See Figure 3.

Fig. 3. Personal self-efficacy (low\_high)
