Aspect of a Personality- Agreeableness - Research Proposal Example

DESIGNING ITEMS TO ACCURATELY MEASURE AN ASPECT OF PERSONALITY - AGREEABLENESS INTRODUCTION Personality has been an important factor in one’s life. It is defined as the combination of the mental, physical, emotional and social characteristics of an individual. Measuring personality places a significant role in the defining the behavioral characteristics of the individual. This is accomplished by measuring the five words related to personality: Openness (curiosity), Extraversion (positive emotions), Conscientiousness (self-discipline), Neuroticism (emotional instability) and Agreeableness (compassionate) also called as the Five Factor Model. Measuring personality based on the five-factor model speculates Agreeableness as a significant dimension of inter-individual dispositional disparity (Goldberg, 1990; Costa & McCrae, 1992) High Agreeableness is found to be associated with friendliness, warmth and compliance to others’ needs (Graziano & Eisenberg, 1997; Takemoto-Chock & Digman, 1981). It is uniquely predictive of harmonious relationships and social support (Soldz & Vaillant, 1999; Asendorpf & Wilpers, 1998) and is negatively associated with aggression, anger and interpersonal arguments (Meier & Robinson, 2004; Jensen-Campbell & Graziano, 2001). Experience-sampling studies show that high-Agreeable adults are more anxious and react in a different way to interpersonal situations than low-Agreeable adults (Cote & Moskowitz, 1998). The high-Agreeable individuals are found to be more concerned in maintaining pleasant relationships and are bothered about potential real interpersonal conflicts (William Damon, Richard M). More Agreeable youths and adults conquer the generation of conflicts and are found to have greater ability to handle interpersonal conflicts that arise. On research, it was found that Agreeableness was measured only as part of the big five factor model. No specific research was performed to study the factor independently. No dedicated questionnaire was found to study this trait. Even the popular Goldberg’s big five personality questionnaire included some items measuring Agreeableness. A questionnaire is a repeated measures design that is made up of multiple items each of which elicits a response from the same person. With the objective to design items to accurately measure an aspect of personality, a questionnaire was developed with the items which adequately measured the personality factor of Agreeableness. Paul Kline’s suggestions to develop items for a scale were followed to obtain a good psychometric score. The most preferable closed question format was adopted to enable the respondents to answer the short question or statement from a number of options. Such procedure enables quantification of the answers into data that provide the required information. The International Personality Item Pool (IPIP) website provides access to a number of personality and individual difference measures. The site provides freely available versions of all the main personality measures as well as over 200 measures of personality individual difference traits. The Goldberg’s personality questionnaire was adopted as baseline to prepare the questionnaire to measure Agreeableness. A list of 50 items was prepared to measure Agreeableness. Both positive and negative questions were framed to measure the factor. The questionnaire used a 5-point Likert scale to measure the how accurately each statement describes the respondent. METHODOLOGY The data was collected from 20 undergraduate Psychology students recruited randomly via the EPR system. The students were given the reason for the research and the details of their participation along with the assurance of confidentiality. A consent statement was obtained from them along with their email ids for further communication purposes. The questionnaire contained details about answering the questions which would explain the respondent’s current behavior. Among the 50 items, 25 items rated the positiveness of the trait Agreeableness while the remaining items rated the trait in a negative manner. The 5-point Likert scale used in the questionnaire was Very Inaccurate, Moderately Accurate, Neither Inaccurate nor Accurate, Moderately Accurate and Very Accurate. The collected data was loaded on to the SPSS (Statistical Package for Social Science) software for analysis. RELIABILITY In psychometric testing involving questionnaire, it is essentially important to test the reliability of the questionnaire. A measure of the internal reliability of the questions helps us to report that all the questions or items are measuring the same thing (here Agreeableness). A correlation between all the individual questions on Agreeableness scale should suggest the questions go together to form a single construct of Agreeableness. A common statistical technique used to measure the internal reliability is Cronbach’s alpha, denoted by α. It is used to assess the level of internal reliability that a set of items has. The level ranges from -1 to +1. Generally, a level of 0.7 and above assures good internal reliability, however, under certain circumstances a much higher level is desired. The other form of reliability is the test-retest reliability. It assesses reliability over time. It is used when researchers studying a particular construct are interested in evaluating the relative consistency of the respondents in their attitudes and behavior over time. FACTOR ANALYSIS Factor analysis is a multivariate (involving multiple variables) data reduction statistical technique that allows us to simplify the correlations between a number of variables. In other words, it can be used to establish that several tests measure the same factor, thereby giving reasons for performing fewer tests. Based on the highest correlations of the original variables with principal factors, it is used to choose a smaller set (subset) of variables from a larger set. It is frequently used to develop questionnaires. The Exploratory Factor Analysis consists of two steps. It first provides the number of factors that exist for the variables under study. This is called Extraction of Factors. The next step called the Rotation of Factors determines which variable loads on which factor. Based on the loadings which range from -1 to +1, the significance of the variable to the factor is determined. That is, higher the number is on the factor, the more important the variable is to the factor. Kline suggested that any number less than 0.3 has to be ignored. EXTRACTION OF FACTORS - PRINCIPAL COMPONENT ANALYSIS The most common Extraction procedure is called the Principal Component Analysis (PCA). Here component means factor. The PCA is used to simplify the correlations between the variables. The PCA forms a linear combination of variables under study such that the largest variance is extracted from the variables. This variance is then removed and second linear combination which explains the maximum proportion of the remaining variance and so on. As a result uncorrelated factors are obtained. PCA thus analyzes the total variance including common and unique variances. In SPSS, the total variance is explained by the eigenvalues. The Eigenvalues above 1 denote the number of factors extracted. This is a traditional way of deciding the factors. As these values are claimed to be unreliable, the Scree Test is used to determine the factors. The Scree Test plots the eigenvalues which provides a visual assessment of which factors has to be extracted. We determine the number of factors by selecting those eigenvalues that occur before the plot straightens out. Another way is to select all points above the elbow in the plot. The number of points denotes the number of factors required to be extracted. ROTATION OF FACTORS Having extracted the factors, it is necessary to know how the extracted factors are related. The rotation of factors provides this idea and also gives a clear picture of which items or variables load on which factor. There are two types of rotation techniques: Orthogonal rotation and Oblique rotation. The Orthogonal rotations assume that every factor is unique to the other. Hence this procedure is preferably used in psychology where a theoretical model under study predicts that the factors are independent. The most often used rotation procedure within the orthogonal rotation is called the Varimax rotation. All these procedures can be easily performed using the SPSS software. CORRELATION Before performing the factor analysis of the items in the questionnaire, it is highly essential to study the correlation between the items. That is, if the test questions measure the same underlying dimension(s), then they would be correlated with each other. However, high correlation among the variables also disturbs the analysis by way of multicollinearity. If any variable is found to have no significant or highly significant correlation with the other variables, then the variable has to be excluded from the study. The correlation matrix provides the correlation coefficients of all the variables in the study. Multicollinearity can be detected from the determinant of the R-matrix in the SPSS output. RESULTS The 50 questions and the responses of the 20 students were loaded onto the SPSS file. Before performing the desired statistical analysis, the internal reliability of the questions was measured. The Cronbach’s alpha was the reliability statistic. The value was found to be 0.87. Since this was a significant number, the questionnaire was considered to be acceptatble in terms of internal reliability. The reliability analysis was again used to select the items using the software. This is possible by selecting the required options in the Statistics section. The Item-Total Statistics table generated provided the Correlation and Cronbach’s alpha value (if item deleted) for all the 50 items. All items with corrected item-total correlation of 0.20 and below were considered to be poorly performing questions. 13 questions were found to have poor correlations: Q4 = -0.07, Q8 = 0.19, Q13 = -0.27, Q17 = -0.35, Q23 = 0.13, Q29 = 0.19, Q31 = -0.03, Q32 = 0.06, Q34 =0.07, Q40 = 0.13, Q41 = 0.14, Q42 = 0.07 and Q46 = -0.11. Deleting these questions would serve to increase the Cronbach’s alpha. The analysis was re-run and new Cronbach’s alpha for the remaining 37 items was obtained as 0.93. However, the correlations of four questions was found to be below the desired correlation of 0.30; Q6 = 0.25, Q14 = 0.25, Q22 = 0.28 & Q36 = 0.22. Yet removing these and recalculating Cronbach’s alpha resulted in the same alpha score. After deciding the questions to be involved in the study, the Factor Analysis option was selected. The descriptive options chosen were Co-efficients, Significance levels and Determinant. These options provided the R-matrix along the significance value of each correlation in the R-matrix. The Determinant of this matrix at the bottom of the matrix was essential to test for multicollinearity or singularity. The Principal Component Analysis method was chosen for the extraction of factors. The Scree plot was chosen for the Display of factor extraction. In the Extract option, Eigenvalue over 1 was decided. From the R-matrix obtained for the 50 items or questions involved in the study, the 5th and 19th questions were found to be highly correlated with other (r = 0.903). The Determinant was found to be 0.000. As this value is less than the required value of 0.00001, to reduce multicollinearity among the items, the two questions were removed and analysis was re-run. Again a low determinant was obtained which showed the existence of multicollinearity. The following Scree Plot was generated. A study of the Scree plot and the Eigenvalues over 1 suggested the extraction of same number of factors. They showed the extraction of 14 factors from among the 50 questions. However the extraction of 14 factors was considered to be too high. Hence it was decided to choose the factors based on the percentage of the total variance explained by the factors. Table 1 provides the Initial Eigenvalues, Extraction Sums of Squared Loadings and Rotated Sums of Squared Loadings of the first 20 factors. As the remaining factors were not extracted, they have been avoided. The percentage of variance given by the Extraction Sums of Squared Loading shows that 24.489% of the variance was explained by factor 1 while 14.572% and 8.932% of the variance were explained by factors 2 and 3 respectively. On rotation, these factors accounted for only 6.013%, 5.426% and 3.594% of the total variance respectively. Thus the first three factors were identified as the most important factors for the study. Table 1: Total Variance Explained Component Initial Eigenvalues Extraction Sums of Squared Loadings Rotation Sums of Squared Loadingsa Total % of Variance Cumulative % Total % of Variance Cumulative % Total 1 11.754 24.489 24.489 11.754 24.489 24.489 6.013 2 6.994 14.572 39.060 6.994 14.572 39.060 5.426 3 4.288 8.932 47.992 4.288 8.932 47.992 3.594 4 3.835 7.990 55.982 3.835 7.990 55.982 2.970 5 3.030 6.312 62.294 3.030 6.312 62.294 3.808 6 2.613 5.443 67.737 2.613 5.443 67.737 4.065 7 2.526 5.262 72.999 2.526 5.262 72.999 4.861 8 2.019 4.206 77.205 2.019 4.206 77.205 3.918 9 1.839 3.831 81.036 1.839 3.831 81.036 4.582 10 1.735 3.614 84.651 1.735 3.614 84.651 4.509 11 1.593 3.319 87.970 1.593 3.319 87.970 2.264 12 1.289 2.685 90.655 1.289 2.685 90.655 5.225 13 1.220 2.541 93.196 1.220 2.541 93.196 5.390 14 1.050 2.187 95.383 1.050 2.187 95.383 4.158 15 .682 1.422 96.805 16 .535 1.115 97.920 17 .439 .915 98.835 18 .348 .726 99.560 19 .211 .440 100.000 20 1.955E-15 4.073E-15 100.000 Extraction Method: Principal Component Analysis. a. When components are correlated, sums of squared loadings cannot be added to obtain a total variance. The Factor Analysis was re-run in SPSS by selecting the number of factors = 3 in the Extract option. Since the factors have been identified, the Component Matrix (Table 2) provides the loadings of each question on each factor. The order of loading enables us to identify any common themes among the questions in each of the 3 factors. All loadings below 0.4 have been suppressed and hence not displayed. Table 2: Component Matrixa Component 1 2 3 q33 .799 q44 .787 q39 .777 q25 .770 q49 .737 q3 .726 q11 .716 .535 q47 .712 q24 .693 q26 .685 q50 .636 q30 .634 q10 .623 q28 .614 -.402 q27 .602 q48 .594 q38 .587 .516 q18 .580 q15 .527 q7 .526 q1 .512 q45 .496 q17 -.480 q22 .402 q16 q14 q23 q43 .757 q6 .754 q29 .731 q41 .717 q37 .684 q2 .680 q42 .677 q20 .521 .550 q40 -.535 q32 .507 q31 .485 q34 .453 q21 .419 .448 q46 q9 .777 q35 .461 -.601 q4 .598 q12 .586 q36 .469 q8 q13 Extraction Method: Principal Component Analysis. a. 3 components extracted. The questions that loaded highly on factor 1 seemed to relate to measuring the importance of inter-personal relationships. People probably scored high on this factor to have high Agreeableness. The questions that loaded highly on factor 2 were largely reversed questions relating to measuring negative personality traits. People probably scored very low on this factor to have high Agreeableness (i.e. low on the negative traits). The questions that loaded highly on factor 3 were related to measuring identification of friendly characteristics. People probably scored high on the positive traits of this factor to have high Agreeableness. DISCUSSION Thus Agreeableness appears to be constructed from the three main factors which are belief in the importance of inter-personal relationships, agreement on negative personality traits (don’t have them or don’t like them) and the identification of friendly characteristics. Further it should also be noted that only the first and second factors account for most of the variance in the data i.e. 24% and 14% respectively. The third factor accounted for only 8% roughly. However, three factors were chosen because they accounted for the highest variance together. The final questionnaire comprised only 48 questions of which 27 questions measured inter-personal relationships, 14 of them measured the factor negatively. 15 questions measured personality traits, 12 of them measured the factor negatively and only 6 questions measured friendly characteristics. Thus inter-personal relationships and personality traits played a significant role in measuring Agreeableness. This was not possible in the available personality questionnaire where the trait is measured along with a few more traits. On a future research process, the predominant quality of the individual contributing to the inter-personal relationship factor and personality trait can be measured. It should also be noted that multicollinearity severely existed among the questions under study despite removing the highly correlated questions. It has to be understood that multicollinearity does not reduce the reliability of the questionnaire or model. It only affects the calculations regarding the variables or questions under study. Thus this affects the analysis of the questionnaire. Moreover in SPSS, significant results are not obtained when multicollinearity exists. Hence the highly correlating questions are removed with the intention to reduce multicollinearity among items. The questions finally obtained from the factor analysis can be re-examined or re-written in some cases to reduce the multicollinearity. REFERENCES Asendorpf, J. B., & Wilpers, S. (1998). Personality effects on social relationships. Journal of Personality and Social Psychology Costa, P. T., & McCrae, R. R. (1992). Four ways five factors are basic. Personality and Individual Differences Daniel Nettle and Bethany Liddle (2008). Agreeableness is related to Social-cognitive but not social-perceptual theory of Mind. European Journal of Personality. David Garson G. Factor Analysis: Statnotes. North Carolina University. http://faculty.chass.ncsu.edu/garson/PA765/factor.htm Field (2005). Research Methods II: Factor analysis on SPSS Goldberg, L. R. (1990). An alternative ‘description of personality’: The Big-Five factor structure. Journal of Personality and Social Psychology Graziano, W. G., & Eisenberg, N. H. (1997). Agreeableness: A dimension of personality. Handbook of personality psychology. International Personality Item Pool. http://ipip.ori.org/ipip/ Jensen-Campbell, L. A., & Graziano, W. G. (2001). Agreeableness as a moderator of interpersonal conflict. Journal of Personality. John Maltby (2009). Psychometric Testing: Factor Analysis, Reliability and Validity Meier, B. P., & Robinson, M. D. (2004). Does quick to blame mean quick to anger? The role of Agreeableness in dissociating blame and anger. Personality and Social Psychology Bulletin. Oppenheim, A.N. (1992) Questionnaire design, interviewing, and attitude measurement. London; Pinter. Rosenthal,R. & Rosnow,R.L.(1991). Essentials of behavioral research. New York: McGraw-Hill. Science Daily (Aug. 11, 2006). Personality Predictors of Intelligence change from younger to older adulthood. Soldz, S., & Vaillant, G. E. (1999). The big five personality traits and the life course: A 45-year longitudinal study. Journal of Research in Personality Takemoto-Chock N. K. & Digman, J. (1981). Factors in the natural language of personality: Re-analysis, comparison and interpretation of six major studies. Multivariate Behavioral Research. William Damon, Richard M. Lerner, Nancy Eisenberg (2006). Temperament and Personality Traits: A Process- Focused, Developmental Taxonomy from Childhood to Adulthood. Handbook of Child Psychology Read More