# An Investigation of Holland Types and the Sixteen Personality Factor Questionnaire--Fifth Edition.

The authors investigated the stability of the published Sixteen
Personality Factor Questionnaire--Fifth Edition (16PF; S. Conn & M.
Rieke, 1994) predictors in predicting Holland types as measured by the
Self-Directed Search (SDS; J. Holland, B. Fritzsche, & A. Powell,
1994). Because the majority of the published regression equations contained unstable predictors, the authors developed modified multiple
regression equations using the more stable predictors. However, these
equations, although statistically significant, shared less than 50% of
the variance with the criterion variable, suggesting limited practical
utility or support for the domain overlap of the 16PF and the SDS. The
authors recommend that the SDS be used when a measure of SDS types is
needed.

The Sixteen Personality Factor Questionnaire--Fifth Edition (16PF; Conn & Rieke, 1994) is consistently rated as one of the most used and researched personality tests (Cattell, Eber, &Tatsuoka, 1970; Walsh & Betz, 1995). It contains 16 bipolar scales (called "primary factors") and several validity scales, with 15 of the factors measuring personality traits and 1 factor measuring cognitive ability or reasoning ability (Conn & Rieke, 1994). One reason the 16PF has been such a popular measure is that validated special scores greatly expand the utility of the 16PF for the counselor. These scores allow the instrument to assess the role of personality structure in leadership, creativity, and specific occupations. Thus, the instrument not only allows the client's interests and abilities to be examined but also allows his or her personality to be taken into consideration during occupational decision making. For example, a client may have interests that are similar to those of a surgeon but may have a score on the 16PF that indicates a great degree of impulsivity and impatience. Of course, this characteristic would need to be addressed during the career decision-making process with this client. Although certainly not all clients would benefit from such a discussion, many occupations (e.g., police officer, clergy, or airline pilot) do require that the personality of the applicant be taken into consideration.

One set of special scores obtained from the 16PF, available by computer scoring, is the prediction of Holland's occupational types. The intent of these special scores is to allow the career counselor to explore the client's interests and personality structure in the career counseling process (Conn & Rieke, 1994). Using the 16PF and the Self-Directed Search (SDS; Holland, Fritzsche, & Powell, 1994) in career counseling requires understanding the overlap of personality and interests, in general, and the ability of the 16PF to predict the SDS codes, in particular.

The field of career counseling continues to examine the relationship between personality and interests and to debate whether there is an overlap of personality and interests or whether these constructs are largely separate domains (Holland, Fritzsche, et al., 1994; Janda, 1998; Oliver, Lent, & Zack, 1998; Young & Chen, 1999; Zunker, 1994). Because the 16PF is one of the most commonly used personality measures that has application to career counseling, the overlap of the 16PF and interests may be of particular importance to career counselors (Oliver et al., 1998; Young & Chen, 1999; Zunker, 1994). The proposed overlap of interests may be important for several reasons, but arguably one of the more important reasons for the career counselor may be the utility of making assumptions about the personality of the client from interest inventory results or of making assumptions about interests given a client's personality structure. If empirical support for the overlap of the personality and interest domains can be de monstrated, then the career counselor may be able to discuss the client's personality characteristics that may be important for a specific career, given the results of an interest inventory. Without the establishment of empirical support for the overlap of these domains, making any assumptions about personality from interests, or interests from personality, is risky, at best.

An additional possible benefit of empirical support for the overlap of interests and personality domains, especially in today's HMO-styled market where assessment time is often limited, is a reduction in testing time in situations when information about personality and interests is beneficial. It can certainly be argued that not all career decisions would benefit from data on personality and interests; however, for situations in which it would be beneficial, a substantial reduction in resources may be possible. The Holland types were likely choices to examine the overlap of more general personality and occupational types because Holland conceptualized these occupational types as "personality types" (Holland, Fritzsche, et al., 1994, p. 1). He contended that there are six occupational personality types, found in both people and the environment, that can be described using a hexagonal model (Holland, Fritzsche, et al., 1994). Holland further contended that the better the match between a person's personality typ e and his or her work environment, the more likely the individual is to find the occupation satisfying. Because Holland conceived of his model as a personality model, we believed that it seemed logical to examine the overlap of interests and personality traits.

The authors of the 16PF have attempted to examine the possible overlap of the personality domain with interests (Conn & Rieke, 1994). In doing so, they used 16PF scores to predict Holland types that would be obtained from the SDS (Karol, 1994). In the development of regression equations, one of the most critical aspects is the cross-validation process. Cross-validation demonstrates that the predictor variables are stable across various samples from the target population. The stability of the predictors is vital to any use of the derived scores, because the applicability of the equations beyond the sample used to develop them is not known without cross-validation research (Hair, Anderson, Tatham, & Black, 1998; Nunnally & Bernstein, 1994). Despite the recommendation for continued cross-validation of regression equations, to date, limited data on the cross-validation of these equations are available (Gonn & Rieke, 1994; Karol, 1994). Therefore, additional cross-validation research regarding these special scores is needed.

The primary purpose of this study was to investigate the stability of the published 16PF predictor variables used in the regression equations to estimate SDS Holland types when a different sample of adults was surveyed. A secondary purpose of this study was to examine the domain overlap of the 16PF and SDS.

Method

Participants

The sample used in this study consisted of the responses from 234 volunteers. Of these volunteers, 109 (47%) were from southern Indiana, southern Illinois, or northern Kentucky; 59 (25%) were from northeastern Ohio or northwestern Pennsylvania; and 65 (28%) were from southern Florida. Participants ranged in age from 18 to 69 years old. One person was not clear regarding location. The mean age was 27.9 years (SD = 10).

Of the participants, 29% (n = 67) were men and 71% (n = 167) were women. To obtain as representative a sample as possible, minority community and university organizations in several locations were contacted and their members were invited to participate. The reported racial heritages of the participants were 83% Whites (n = 194), 9% Blacks (n = 20), 6% Hispanics (n = 14), 2% Asians (n = 5), and 0.4% Native American (n = 1). The participants' mean educational level was 14.9 years (2 years of college; SD = 2.5). One participant did not report his or her educational level.

There were 58% (n = 135) undergraduate and graduate students and 42% (n = 98) community members (persons not enrolled in classes) in the sample. One participant's group membership was not indicated. To expand the diversity of interests in the sample, university participants were solicited from programs in art, music, accounting, business management, counseling, elementary education, secondary education, and special education.

Instruments

The 16PF. The 16PF is designed to measure normal personality traits (Gattell et al., 1970; Conn & Rieke, 1994). It has been revised several times over the years, primarily to update norms, but more recently to update language and to improve the psychometric qualities of the tool. The 16PF contains 16 bipolar scales (called "primary factors"), 5 global factor scales, and several validity scales. Fifteen of the primary factors and the 5 global factors measure personality traits; 1 factor measures cognitive ability. The stability coefficients for the personality factors and validity scales range from r = .69 to r .91 for 2 weeks, with the range of r = .56 to r = .82 for 2 months. The internal consistency of the primary factors and validity scales ranged from [alpha] = .66 to [alpha] = .87 (Conn & Rieke, 1994).

There are extensive supportive validity data reported in the technical manual for the 16PF. Data on the factor structure, item analysis, and relationships of the 16PF to other measures are also included in the manual (Conn & Rieke, 1994. Using data from a sample of 194 individuals, Karol (1994) developed equations, using a multiple regression procedure, for predicting SDS Holland codes from 16PF data.

The SDS. The SDS is a widely used interest inventory, developed by Holland, that was designed to aid in the career decision-making process (Holland, Fritzsche, et al., 1994; Holland, Powell, & Fritzsche, 1994; Walsh & Betz, 1995). The SDS has six scales that are based on Holland's theory of career decision making. Holland described the six scales using a hexagonal model that symbolizes the relationship between the personality types (Holland, Fritzsche, et al., 1994; Holland, Powell, et al., 1994).

The stability of the six SDS scales ranged from r = .76 to r= .89 over 4 to 12 weeks. Considerable validity data are reported in the manuals, including examinations of the content, relationship to other interest measures, relationship to personality measures, and prediction of occupational satisfactions and fit (Holland, Fritzsche, et al., 1994; Holland, Powell, et al., 1994).

Procedure

Data for this study were collected as part of a group of studies on the 16PF being conducted by the first author. Sample size was preestablished, using procedures recommended by experts in the application of this methodology (Hair et al., 1998; Nunnally & Bernstein, 1994).

The administration packets included the 16PF, SDS, demographics sheet, and other instruments, which were placed in an envelope in counterbalanced order. A description of the study was either handed to or read to participants before they opened the envelope. Approximately 60% of the participants completed the inventory in a group setting, and the other 40% took the packet home to complete it. Two hundred and sixty packets of materials were distributed; 234 were returned with completed 16PFs and SDSs, yielding a return rate of 90%.

Data Analysis

Data analysis was performed using SYSTAT 9.0 for Windows (SPSS, 1999). The 16PF raw scale scores were converted to sten scores, using the combined population norms provided by the publisher (Conn & Rieke, 1994).

Because this study was designed to examine the effect of sampling error (or bias) on the stability of predictors in multiple regression equations, an established procedure to detect error in regression formulas was applied; the procedure was first developed by Cleary (1968) and further refined by others (Hair et al., 1998; Nunnally & Bernstein, 1994). Sampling bias (error) is suggested if the predictors proposed to make up the regression equation do not meet two criteria in the cross-validation sample. The first criterion is that the predictor variable must significantly enter into the regression equation when using the cross-validation sample. Second, the predictor variable must enter into the regression equation from the cross-validation sample in the same direction (positive or negative) as in the developmental sample (Hair et al., 1998; Nunnally & Bernstein, 1994).

Two regression procedures (models) were calculated to evaluate whether the predictor variables that were reported by Karol (1994) met the two cross-validation criteria. The first regression model (free model) was calculated using a forward stepwise procedure with the entry and removal criterion set to .10 as was done during development (Karol, 1994). Stability of the predictors was supported if the predictor entered into the free model as it had in Karol's study. The second regression model (forced model) was constructed by using the published predictors for each Holland type to calculate a multiple regression equation. This forced all the prior established predictors into a model to determine whether, if they had entered into the equation, they would have done so in the same direction, using data from the sample of individuals who participated in this study. Correct entry into the forced model provided less support for the stability of the predictors than did entry into the free model. Stability of the predi ctors was best supported if the predictor variables entered into the free model equation and did so in the same direction as found in Karol's study (Hair et al., 1998; Nunnally & Bernstein, 1994).

A cross-validated equation was calculated using predictors that were considered stable from Karol's (1994) study and the free model. These represent the cross-validated models (Cleary, 1968; Hair et al., 1998; Nunnally & Bernstein, 1994). The cross-validated model was considered statistically significant if it provided a prediction that was substantially greater than chance, as determined by the Fstatistic having a p < .05.

Domain overlap can be considered an issue of practical significance because two tools or procedures are attempting to measure a common construct or domain in regression models (Nunnally & Bernstein, 1994). The practical utility of two measures being considered interchangeable has been examined using a number of methods, but most commonly by examining alternate form reliability (Cicchetti, 1994; Janda, 1998; Nunnally & Bernstein, 1994). Two tools or procedures are accepted as sharing adequate domain, or seen as alternate forms, if they have 50% or more shared variance. This provides initial evidence of domain overlap. The shared variance of regression models is best estimated by examining the adjusted [R.sup.2]. The cross-validated multiple regression equations were considered practically significant if they shared 50% or more variance (adjusted [R.sup.2] > .50) with the SDS scale they were predicating.

Results

The means and standard deviations for the 16PF and the SDS are provided in Table 1. The results of examining the stability of the predictors from the primary and global factor models are summarized in Table 2. These data supported the stability of the global factor predictors for the SDS Realistic, Artistic, and Conventional scales. These data supported the primary factor model for the SDS Enterprising scale as being composed of stable predictors.

Table 3 presents the adjusted [R.sup.2]s for both the 16PF global factor and 16PF primary factor cross-validated multiple regression equations. None of the multiple regression equations met the criteria for practical significance in predicting the SDS scales from 16PF factors of adjusted [R.sup.2] > .50.

Discussion

The primary purpose of this study was to investigate, using a different sample of individuals than had been used in previous studies, the stability of the 16PF predictor variables in predicting Holland types, as measured by the SDS. If the stability of the predictor variable could not be demonstrated, the application of the equation in a real-life situation would not be supported. This would mean that any equation would be different for various groups of adults and that no consistent prediction of SDS types would be possible.

The results of this study indicate that the published global factor multiple regression equations for the Realistic, Artistic, and Conventional scales were constructed from stable predictor variables. Three of the published global factor multiple regression equations contained predictors that failed to be supported as stable in this study. Of the six published multiple regression equations using the l6PF primary factors to predict SDS scores, only the Enterprising equation was found to be constructed completely from stable predictors. These data indicate that if a prediction of SDS scale scores from the 16PF is needed, the cross-validated multiple regression equations reported in Table 3 contain predictors known to be more stable, and, as such, are preferable to Karol's (1994) equations. Because there were statistically significant regression equations that could be constructed from the 16PF, this provided some support for the potential of the 16PF to have utility in exploring SDS types. However, before such a potential can be considered meaningful and as having practical utility in career counseling, the establishment of alternate form reliability also needs to support such a procedure. Furthermore, these results suggest the need to continue to refine the regression models using other samples and further research into the relationship of the Holland types and the 16PF.

In order to examine the practical utility of the regression equations for the SDS that were constructed from the 16PF, one method is to examine adequate alternate form reliability. When the alternate form reliability of the two procedures was examined, no support was found for the interchangeability of the two procedures. The results of this study suggest that statistically significant and stable regression models that predict SDS scale scores can be constructed from 16PF scores. This finding provides evidence that there is some shared domain between the SDS and the 16PF. However, neither Karol's (1994) results nor the results from this study provided support for the practical significance of any model. Thus, although there is likely some shared domain between the two measures, it is insufficient to equate them as alternate forms for one another. This suggests that although the overlap may be of interest to researchers, the application of these equations by the career counselor in a real-life setting was not supported by these data. Thus, when measures of interests and personality are important to the career decision-making process, both the 16PF and the SDS would need to be administered.

To ensure that the lack of utility of the 16PF in measuring SDS scores was not an artifact of the statistical procedure of multiple regression, classical item analysis was used to examine these data. These results further supported the finding that although some degree of content overlap was present, it was insufficient to have practical utility (Nunnally & Bernstein, 1994). The lack of domain overlap was most evident in the prediction of the SDS Conventional scale score. There was less than 5% shared variance between the cross-validated multiple regression equations and the SDS Conventional scale score. The low predictive power of the cross-validated multiple regression was consistent with the general lower predictive power of the SDS Conventional scale in this study and with the results reported by Karol (1994).

In summary, these results have implications for the practitioner and for future researchers. For the practitioner, these results suggest that although the 16PF may have some domain overlap with the SDS, the overlap is too small to be of utility in an applied setting.

Researchers may note, however, that cross-validated statistically significant regression equations could be constructed. This suggests that at least some of the domain of personality, as measured by the 16PF and the SDS, was present. These data were consistent with Karol's (1994) findings that the Realistic, Enterprising, Social, and Artistic types had the largest overlap with personality, whereas the Conventional type had the least overlap. This suggests that although some Holland types may have overlap with 16PF personality traits, others may have little overlap. Additional research should examine the possibility that, perhaps as we have defined personality and interests, the overlap is only in some areas.

This study was limited in several ways. Although this sample was more representative of the general population of the United States than the sample used in previous research, the study needs to be replicated using more representative samples. Furthermore, it is important to remember that as with any assessment tool, application and interpretation of the 16PF and SDS need to take place on the basis of sound clinical practice and an understanding of assessment.

Dale R. Pietrzak is an assistant professor in the Department of Counseling and Psychology in Education at the University of South Dakota, Vermillion. Betsy J. Page is an assistant professor in the Department of Counseling and Human Development Services at Kent State University, Kent, Ohio. Correspondence regarding this article should he sent to Dale R. Pietrzak, University of South Dakota, 414 East Clark Street, Vermillion, SD 57069 (e-mail: dpietrza@usd.edu).

References

Catrell, R., Eber, H., & Tatsuoka, M. (1970). Handbook for the J6PF. Champaign, IL: Institute for Personality and Ability Assessment.

Cicchetti, D. V. (1994). Guidelines, criteria, and rules of thumb For evaluating normed and standardized assessment instruments in psychology. Psychological Assessment, 6, 284-290.

Cleary, A. (1968). Test bias: Prediction of grades of Negro and White students in integrated colleges. Journal of Educational Measurement, 10, 43-56.

Conn, S., & Rieke, M. (Eds.). (1994). l6PFfifth edition: Technical manual. Champaign, IL: Institute for Personality and Ability Testing.

Hair, J., Anderson, R., Tatham, R., & Black, W. (1998). Multivariate data analysis (5th ed.). Upper Saddle River, NJ: Prentice Hall.

Holland, J., Fritzsche, B., & Powell, A. (1994). Self-Directed Search: Technical manual. Odessa, FL: Psychological Assessment Resources.

Holland, J., Powell, A, & Fritzsche, B. (1994). Self-Directed Search: Professional user's guide. Odessa, FL: Psychological Assessment Resources.

Janda, L. (1998). Psychological testing: Theory and application. Boston: Allyn & Bacon.

Karol, D. (1994). Holland occupational typology and the 16PF. In S. Conn & M. Rieke (Eds.), l6PF fifth edition: Technical manual (pp. 2 13-235). Champaign, IL: Institute for Personality and Ability Testing.

Nunnally, J., & Bernstein, I. (1994). Psychometric theory (3rd ed.). New York: McGraw-Hill. Oliver, L., Lent, E., & Zack, J. (1998). Career and vocational assessment 1995-1996: A biennial review. Journal of Career Assessment, 6, 231-268. SPSS. (1999). SYSTAT 9: Statistics I. Chicago: Author.

Walsh, W., & Betz, N. (1995). Tests and assessment (3rd ed.). Englewood Cliffs, NJ: Prentice Hall.

Young, R., & Chen, C. (1999). Annual review: Practice and research in career counseling and development--1998. The Career Development Quarterly, 48, 98-141.

Zunker, V. (1994). Career counseling: Applied concepts of life planning (4th ed.). Pacific Grove, CA: Brooks/Cole.

The Sixteen Personality Factor Questionnaire--Fifth Edition (16PF; Conn & Rieke, 1994) is consistently rated as one of the most used and researched personality tests (Cattell, Eber, &Tatsuoka, 1970; Walsh & Betz, 1995). It contains 16 bipolar scales (called "primary factors") and several validity scales, with 15 of the factors measuring personality traits and 1 factor measuring cognitive ability or reasoning ability (Conn & Rieke, 1994). One reason the 16PF has been such a popular measure is that validated special scores greatly expand the utility of the 16PF for the counselor. These scores allow the instrument to assess the role of personality structure in leadership, creativity, and specific occupations. Thus, the instrument not only allows the client's interests and abilities to be examined but also allows his or her personality to be taken into consideration during occupational decision making. For example, a client may have interests that are similar to those of a surgeon but may have a score on the 16PF that indicates a great degree of impulsivity and impatience. Of course, this characteristic would need to be addressed during the career decision-making process with this client. Although certainly not all clients would benefit from such a discussion, many occupations (e.g., police officer, clergy, or airline pilot) do require that the personality of the applicant be taken into consideration.

One set of special scores obtained from the 16PF, available by computer scoring, is the prediction of Holland's occupational types. The intent of these special scores is to allow the career counselor to explore the client's interests and personality structure in the career counseling process (Conn & Rieke, 1994). Using the 16PF and the Self-Directed Search (SDS; Holland, Fritzsche, & Powell, 1994) in career counseling requires understanding the overlap of personality and interests, in general, and the ability of the 16PF to predict the SDS codes, in particular.

The field of career counseling continues to examine the relationship between personality and interests and to debate whether there is an overlap of personality and interests or whether these constructs are largely separate domains (Holland, Fritzsche, et al., 1994; Janda, 1998; Oliver, Lent, & Zack, 1998; Young & Chen, 1999; Zunker, 1994). Because the 16PF is one of the most commonly used personality measures that has application to career counseling, the overlap of the 16PF and interests may be of particular importance to career counselors (Oliver et al., 1998; Young & Chen, 1999; Zunker, 1994). The proposed overlap of interests may be important for several reasons, but arguably one of the more important reasons for the career counselor may be the utility of making assumptions about the personality of the client from interest inventory results or of making assumptions about interests given a client's personality structure. If empirical support for the overlap of the personality and interest domains can be de monstrated, then the career counselor may be able to discuss the client's personality characteristics that may be important for a specific career, given the results of an interest inventory. Without the establishment of empirical support for the overlap of these domains, making any assumptions about personality from interests, or interests from personality, is risky, at best.

An additional possible benefit of empirical support for the overlap of interests and personality domains, especially in today's HMO-styled market where assessment time is often limited, is a reduction in testing time in situations when information about personality and interests is beneficial. It can certainly be argued that not all career decisions would benefit from data on personality and interests; however, for situations in which it would be beneficial, a substantial reduction in resources may be possible. The Holland types were likely choices to examine the overlap of more general personality and occupational types because Holland conceptualized these occupational types as "personality types" (Holland, Fritzsche, et al., 1994, p. 1). He contended that there are six occupational personality types, found in both people and the environment, that can be described using a hexagonal model (Holland, Fritzsche, et al., 1994). Holland further contended that the better the match between a person's personality typ e and his or her work environment, the more likely the individual is to find the occupation satisfying. Because Holland conceived of his model as a personality model, we believed that it seemed logical to examine the overlap of interests and personality traits.

The authors of the 16PF have attempted to examine the possible overlap of the personality domain with interests (Conn & Rieke, 1994). In doing so, they used 16PF scores to predict Holland types that would be obtained from the SDS (Karol, 1994). In the development of regression equations, one of the most critical aspects is the cross-validation process. Cross-validation demonstrates that the predictor variables are stable across various samples from the target population. The stability of the predictors is vital to any use of the derived scores, because the applicability of the equations beyond the sample used to develop them is not known without cross-validation research (Hair, Anderson, Tatham, & Black, 1998; Nunnally & Bernstein, 1994). Despite the recommendation for continued cross-validation of regression equations, to date, limited data on the cross-validation of these equations are available (Gonn & Rieke, 1994; Karol, 1994). Therefore, additional cross-validation research regarding these special scores is needed.

The primary purpose of this study was to investigate the stability of the published 16PF predictor variables used in the regression equations to estimate SDS Holland types when a different sample of adults was surveyed. A secondary purpose of this study was to examine the domain overlap of the 16PF and SDS.

Method

Participants

The sample used in this study consisted of the responses from 234 volunteers. Of these volunteers, 109 (47%) were from southern Indiana, southern Illinois, or northern Kentucky; 59 (25%) were from northeastern Ohio or northwestern Pennsylvania; and 65 (28%) were from southern Florida. Participants ranged in age from 18 to 69 years old. One person was not clear regarding location. The mean age was 27.9 years (SD = 10).

Of the participants, 29% (n = 67) were men and 71% (n = 167) were women. To obtain as representative a sample as possible, minority community and university organizations in several locations were contacted and their members were invited to participate. The reported racial heritages of the participants were 83% Whites (n = 194), 9% Blacks (n = 20), 6% Hispanics (n = 14), 2% Asians (n = 5), and 0.4% Native American (n = 1). The participants' mean educational level was 14.9 years (2 years of college; SD = 2.5). One participant did not report his or her educational level.

There were 58% (n = 135) undergraduate and graduate students and 42% (n = 98) community members (persons not enrolled in classes) in the sample. One participant's group membership was not indicated. To expand the diversity of interests in the sample, university participants were solicited from programs in art, music, accounting, business management, counseling, elementary education, secondary education, and special education.

Instruments

The 16PF. The 16PF is designed to measure normal personality traits (Gattell et al., 1970; Conn & Rieke, 1994). It has been revised several times over the years, primarily to update norms, but more recently to update language and to improve the psychometric qualities of the tool. The 16PF contains 16 bipolar scales (called "primary factors"), 5 global factor scales, and several validity scales. Fifteen of the primary factors and the 5 global factors measure personality traits; 1 factor measures cognitive ability. The stability coefficients for the personality factors and validity scales range from r = .69 to r .91 for 2 weeks, with the range of r = .56 to r = .82 for 2 months. The internal consistency of the primary factors and validity scales ranged from [alpha] = .66 to [alpha] = .87 (Conn & Rieke, 1994).

There are extensive supportive validity data reported in the technical manual for the 16PF. Data on the factor structure, item analysis, and relationships of the 16PF to other measures are also included in the manual (Conn & Rieke, 1994. Using data from a sample of 194 individuals, Karol (1994) developed equations, using a multiple regression procedure, for predicting SDS Holland codes from 16PF data.

The SDS. The SDS is a widely used interest inventory, developed by Holland, that was designed to aid in the career decision-making process (Holland, Fritzsche, et al., 1994; Holland, Powell, & Fritzsche, 1994; Walsh & Betz, 1995). The SDS has six scales that are based on Holland's theory of career decision making. Holland described the six scales using a hexagonal model that symbolizes the relationship between the personality types (Holland, Fritzsche, et al., 1994; Holland, Powell, et al., 1994).

The stability of the six SDS scales ranged from r = .76 to r= .89 over 4 to 12 weeks. Considerable validity data are reported in the manuals, including examinations of the content, relationship to other interest measures, relationship to personality measures, and prediction of occupational satisfactions and fit (Holland, Fritzsche, et al., 1994; Holland, Powell, et al., 1994).

Procedure

Data for this study were collected as part of a group of studies on the 16PF being conducted by the first author. Sample size was preestablished, using procedures recommended by experts in the application of this methodology (Hair et al., 1998; Nunnally & Bernstein, 1994).

The administration packets included the 16PF, SDS, demographics sheet, and other instruments, which were placed in an envelope in counterbalanced order. A description of the study was either handed to or read to participants before they opened the envelope. Approximately 60% of the participants completed the inventory in a group setting, and the other 40% took the packet home to complete it. Two hundred and sixty packets of materials were distributed; 234 were returned with completed 16PFs and SDSs, yielding a return rate of 90%.

Data Analysis

Data analysis was performed using SYSTAT 9.0 for Windows (SPSS, 1999). The 16PF raw scale scores were converted to sten scores, using the combined population norms provided by the publisher (Conn & Rieke, 1994).

Because this study was designed to examine the effect of sampling error (or bias) on the stability of predictors in multiple regression equations, an established procedure to detect error in regression formulas was applied; the procedure was first developed by Cleary (1968) and further refined by others (Hair et al., 1998; Nunnally & Bernstein, 1994). Sampling bias (error) is suggested if the predictors proposed to make up the regression equation do not meet two criteria in the cross-validation sample. The first criterion is that the predictor variable must significantly enter into the regression equation when using the cross-validation sample. Second, the predictor variable must enter into the regression equation from the cross-validation sample in the same direction (positive or negative) as in the developmental sample (Hair et al., 1998; Nunnally & Bernstein, 1994).

Two regression procedures (models) were calculated to evaluate whether the predictor variables that were reported by Karol (1994) met the two cross-validation criteria. The first regression model (free model) was calculated using a forward stepwise procedure with the entry and removal criterion set to .10 as was done during development (Karol, 1994). Stability of the predictors was supported if the predictor entered into the free model as it had in Karol's study. The second regression model (forced model) was constructed by using the published predictors for each Holland type to calculate a multiple regression equation. This forced all the prior established predictors into a model to determine whether, if they had entered into the equation, they would have done so in the same direction, using data from the sample of individuals who participated in this study. Correct entry into the forced model provided less support for the stability of the predictors than did entry into the free model. Stability of the predi ctors was best supported if the predictor variables entered into the free model equation and did so in the same direction as found in Karol's study (Hair et al., 1998; Nunnally & Bernstein, 1994).

A cross-validated equation was calculated using predictors that were considered stable from Karol's (1994) study and the free model. These represent the cross-validated models (Cleary, 1968; Hair et al., 1998; Nunnally & Bernstein, 1994). The cross-validated model was considered statistically significant if it provided a prediction that was substantially greater than chance, as determined by the Fstatistic having a p < .05.

Domain overlap can be considered an issue of practical significance because two tools or procedures are attempting to measure a common construct or domain in regression models (Nunnally & Bernstein, 1994). The practical utility of two measures being considered interchangeable has been examined using a number of methods, but most commonly by examining alternate form reliability (Cicchetti, 1994; Janda, 1998; Nunnally & Bernstein, 1994). Two tools or procedures are accepted as sharing adequate domain, or seen as alternate forms, if they have 50% or more shared variance. This provides initial evidence of domain overlap. The shared variance of regression models is best estimated by examining the adjusted [R.sup.2]. The cross-validated multiple regression equations were considered practically significant if they shared 50% or more variance (adjusted [R.sup.2] > .50) with the SDS scale they were predicating.

Results

The means and standard deviations for the 16PF and the SDS are provided in Table 1. The results of examining the stability of the predictors from the primary and global factor models are summarized in Table 2. These data supported the stability of the global factor predictors for the SDS Realistic, Artistic, and Conventional scales. These data supported the primary factor model for the SDS Enterprising scale as being composed of stable predictors.

Table 3 presents the adjusted [R.sup.2]s for both the 16PF global factor and 16PF primary factor cross-validated multiple regression equations. None of the multiple regression equations met the criteria for practical significance in predicting the SDS scales from 16PF factors of adjusted [R.sup.2] > .50.

Discussion

The primary purpose of this study was to investigate, using a different sample of individuals than had been used in previous studies, the stability of the 16PF predictor variables in predicting Holland types, as measured by the SDS. If the stability of the predictor variable could not be demonstrated, the application of the equation in a real-life situation would not be supported. This would mean that any equation would be different for various groups of adults and that no consistent prediction of SDS types would be possible.

The results of this study indicate that the published global factor multiple regression equations for the Realistic, Artistic, and Conventional scales were constructed from stable predictor variables. Three of the published global factor multiple regression equations contained predictors that failed to be supported as stable in this study. Of the six published multiple regression equations using the l6PF primary factors to predict SDS scores, only the Enterprising equation was found to be constructed completely from stable predictors. These data indicate that if a prediction of SDS scale scores from the 16PF is needed, the cross-validated multiple regression equations reported in Table 3 contain predictors known to be more stable, and, as such, are preferable to Karol's (1994) equations. Because there were statistically significant regression equations that could be constructed from the 16PF, this provided some support for the potential of the 16PF to have utility in exploring SDS types. However, before such a potential can be considered meaningful and as having practical utility in career counseling, the establishment of alternate form reliability also needs to support such a procedure. Furthermore, these results suggest the need to continue to refine the regression models using other samples and further research into the relationship of the Holland types and the 16PF.

In order to examine the practical utility of the regression equations for the SDS that were constructed from the 16PF, one method is to examine adequate alternate form reliability. When the alternate form reliability of the two procedures was examined, no support was found for the interchangeability of the two procedures. The results of this study suggest that statistically significant and stable regression models that predict SDS scale scores can be constructed from 16PF scores. This finding provides evidence that there is some shared domain between the SDS and the 16PF. However, neither Karol's (1994) results nor the results from this study provided support for the practical significance of any model. Thus, although there is likely some shared domain between the two measures, it is insufficient to equate them as alternate forms for one another. This suggests that although the overlap may be of interest to researchers, the application of these equations by the career counselor in a real-life setting was not supported by these data. Thus, when measures of interests and personality are important to the career decision-making process, both the 16PF and the SDS would need to be administered.

To ensure that the lack of utility of the 16PF in measuring SDS scores was not an artifact of the statistical procedure of multiple regression, classical item analysis was used to examine these data. These results further supported the finding that although some degree of content overlap was present, it was insufficient to have practical utility (Nunnally & Bernstein, 1994). The lack of domain overlap was most evident in the prediction of the SDS Conventional scale score. There was less than 5% shared variance between the cross-validated multiple regression equations and the SDS Conventional scale score. The low predictive power of the cross-validated multiple regression was consistent with the general lower predictive power of the SDS Conventional scale in this study and with the results reported by Karol (1994).

In summary, these results have implications for the practitioner and for future researchers. For the practitioner, these results suggest that although the 16PF may have some domain overlap with the SDS, the overlap is too small to be of utility in an applied setting.

Researchers may note, however, that cross-validated statistically significant regression equations could be constructed. This suggests that at least some of the domain of personality, as measured by the 16PF and the SDS, was present. These data were consistent with Karol's (1994) findings that the Realistic, Enterprising, Social, and Artistic types had the largest overlap with personality, whereas the Conventional type had the least overlap. This suggests that although some Holland types may have overlap with 16PF personality traits, others may have little overlap. Additional research should examine the possibility that, perhaps as we have defined personality and interests, the overlap is only in some areas.

This study was limited in several ways. Although this sample was more representative of the general population of the United States than the sample used in previous research, the study needs to be replicated using more representative samples. Furthermore, it is important to remember that as with any assessment tool, application and interpretation of the 16PF and SDS need to take place on the basis of sound clinical practice and an understanding of assessment.

Dale R. Pietrzak is an assistant professor in the Department of Counseling and Psychology in Education at the University of South Dakota, Vermillion. Betsy J. Page is an assistant professor in the Department of Counseling and Human Development Services at Kent State University, Kent, Ohio. Correspondence regarding this article should he sent to Dale R. Pietrzak, University of South Dakota, 414 East Clark Street, Vermillion, SD 57069 (e-mail: dpietrza@usd.edu).

References

Catrell, R., Eber, H., & Tatsuoka, M. (1970). Handbook for the J6PF. Champaign, IL: Institute for Personality and Ability Assessment.

Cicchetti, D. V. (1994). Guidelines, criteria, and rules of thumb For evaluating normed and standardized assessment instruments in psychology. Psychological Assessment, 6, 284-290.

Cleary, A. (1968). Test bias: Prediction of grades of Negro and White students in integrated colleges. Journal of Educational Measurement, 10, 43-56.

Conn, S., & Rieke, M. (Eds.). (1994). l6PFfifth edition: Technical manual. Champaign, IL: Institute for Personality and Ability Testing.

Hair, J., Anderson, R., Tatham, R., & Black, W. (1998). Multivariate data analysis (5th ed.). Upper Saddle River, NJ: Prentice Hall.

Holland, J., Fritzsche, B., & Powell, A. (1994). Self-Directed Search: Technical manual. Odessa, FL: Psychological Assessment Resources.

Holland, J., Powell, A, & Fritzsche, B. (1994). Self-Directed Search: Professional user's guide. Odessa, FL: Psychological Assessment Resources.

Janda, L. (1998). Psychological testing: Theory and application. Boston: Allyn & Bacon.

Karol, D. (1994). Holland occupational typology and the 16PF. In S. Conn & M. Rieke (Eds.), l6PF fifth edition: Technical manual (pp. 2 13-235). Champaign, IL: Institute for Personality and Ability Testing.

Nunnally, J., & Bernstein, I. (1994). Psychometric theory (3rd ed.). New York: McGraw-Hill. Oliver, L., Lent, E., & Zack, J. (1998). Career and vocational assessment 1995-1996: A biennial review. Journal of Career Assessment, 6, 231-268. SPSS. (1999). SYSTAT 9: Statistics I. Chicago: Author.

Walsh, W., & Betz, N. (1995). Tests and assessment (3rd ed.). Englewood Cliffs, NJ: Prentice Hall.

Young, R., & Chen, C. (1999). Annual review: Practice and research in career counseling and development--1998. The Career Development Quarterly, 48, 98-141.

Zunker, V. (1994). Career counseling: Applied concepts of life planning (4th ed.). Pacific Grove, CA: Brooks/Cole.

TABLE 1 Means and Standard Deviations for 16PF-5th Edition and SDS Scale/Factor n M SD 16PF-5th Edition A Warmth 230 6.3 1.8 B Reasoning 229 6.2 1.7 C Emotional stability 230 5.6 1.8 E Dominance 230 5.2 2.0 F Liveliness 230 6.0 1.9 G Rule-consciousness 229 5.4 1.9 H Social boldness 230 5.7 2.0 I Sensitivity 230 6.4 1.8 L Vigilance 230 5.3 2.0 M Abstractness 230 5.6 1.9 N Privateness 230 5.0 2.1 O Apprehension 230 5.8 1.8 Q1 Open to change 230 5.7 2.2 Q2 Self-Reliance 230 5.3 1.8 Q3 Perfectionism 230 5.0 2.1 Q4 Tension 230 5.5 1.8 Extroversion 230 6.1 1.9 Anxiety 230 5.5 2.0 Tough mindedness 230 4.8 2.0 Independence 230 5.4 1.9 Self-Control 229 5.1 1.8 Self-Directed Search Realistic 234 17.2 10.7 Investigative 234 21.9 10.6 Artistic 234 24.2 12.0 Social 234 36.3 8.9 Enterprising 234 25.9 9.0 Conventional 234 21.7 10.8 Note. 16PF-5th Edition = Sixteen Personality Factor Questionnaire-Fifth Edition SDS = Self-Directed Search. For the standardization sample, M = 5.5 and SD = 2.0 TABLE 2 Regression Results of 16PF-5th Edition for SDS Holland Type (a) F R Global Factors Realistic (n = 229) Predicted (c) 10.7 (**) .43 - .47 Free (d) 12.9 (**) .47 Forced (e) 12.9 (**) .47 Investigative (n = 230) Predicted (c) 5.7 (**) .30 - .36 Free (d) 7.2 (**) .30 Forced (e) 4.6 (**) .31 Artistic (n = 230) Predicted (c) 70.9 (**) .51 - .52 Free (d) 38.9 (**) .59 Forced (e) 106.7 (**) .57 Social (n = 229) Predicted (c) 27.2 (**) .53 - .55 Free (d) 23.6 (**) .54 Forced (e) 28.9 (**) .53 Enterprising (n = 230) Predicted (c) 27.0 (**) .52 - .55 Free (d) 24.7 (**) .42 Forced (e) 16.6 (**) .43 Conventional (n = 229) Predicted (c) 19.7 (**) .39 - .41 Free (d) 6.9 (**) .24 Forced (e) 6.9 (**) .24 Primary Factors Realistic (n = 229) Predicted (c) 18.2 (**) .57 - .54 Free (d) 26.8 (**) .61 Forced (e) 25.4 (**) .60 Investigative (n = 229) Predicted (c) 12.8 (**) .57 - .52 Free (d) 11.3 (**) .48 Forced (e) 8.4 (**) .46 Artistic (n = 230) Predicted (c) 21.4 (**) .56 - .53 Free (d) 24.1 (**) .59 Forced (e) 27.7 (**) .57 Social (n = 229) Predicted (c) 21.5 (**) .60 - .57 Free (d) 21.2 (**) .60 Forced (e) 22.7 (**) .58 Enterprising (n = 230) Predicted (c) 20.7 (**) .60 - .57 Free (d) 9.4 (**) .51 Forced (e) 12.1 (**) .46 Conventional (n = 222) Predicted (c) 16.6 (**) .46 - .43 Free (d) 6.9 (**) .29 Forced (e) 5.4 (**) .26 Holland Type (a) 16PF Factors (b) Global Factors Realistic (n = 229) Predicted (c) TM+, IN+, AX-, EX-, SC- Free (d) IN+, TM+, EX-, AX-, SC- Forced (e) IN+, TM+, EX-, AX-, SC- Investigative (n = 230) Predicted (c) EX-, SC-, IN+ (f), TM+, AX- (f) Free (d) EX-, IN+, AX- Forced (e) EX-, IN+, AX-, TM- (g), SC+ (g) Artistic (n = 230) Predicted (c) TM- Free (d) TM-, EX-, IN+ Forced (e) TM- Social (n = 229) Predicted (c) EX+, TM-, SC+ (f) Free (d) TM-, EX+, IN+, SC+ Forced (e) TM-, EX+, SC+ Enterprising (n = 230) Predicted (c) IN+, TM+(f), EX+ Free (d) IN+, TM+ Forced (e) IN+, TM+, EX+ (g) Conventional (n = 229) Predicted (c) SC+, TM+ Free (d) SC+, TM+ Forced (e) SC+, TM+ Primary Factors Realistic (n = 229) Predicted (c) I-, A-, O-, Q1+, Q4- Free (d) I-, Q1+, A-, G+, H+ Forced (e) I-, Q1+, A-, Q4- (g), O- (g) Investigative (n = 229) Predicted (c) A-, I-, B+, Q4-, M+, N- (f), Q1+ Free (d) A-, Q1+, B+, O+, Q4-, I- Forced (e) A-, B+, Q1+, Q4-, I-, N+ (g), M+(g) Artistic (n = 230) Predicted (c) I+, M+, H+, Q1+ Free (d) I+, Q1+, M+, Q2+, E+ Forced (e) I+, Q1+, M+, H+ (g) Social (n = 229) Predicted (c) A+, H+, G+ (f), C- (f), Q1+ Free (d) A+, Q1+, H+, E+, N+, I+ Forced (e) A+, Q1+, H+, C- (g) G+ (g) Enterprising (n = 230) Predicted (c) A+, E+ H+, I-, N+ (f) Free (d) E+, A+, I-, H+, C-, N+, Q3+, Q1+ Forced (e) E+, A+, I-, N+, H+ Conventional (n = 222) Predicted (c) M-, Q3+, I- Free (d) M-, L+, H- Forced (e) Q3+, I-, M- (g) Note. See Table 1 Note. (a)Scales in order of magnitude of Standardized beta weights; + or - = direction scale entered equation. (b)Letters = scale names (Conn & Rieke, 1994). (c)Conn & Rieke (1994). (d)Model developed from sample when forward stepwise procedure and .10 entry/removal criteria used. (e)When predicted scales placed into model and regression equation estimated. (f)Indicates prediction that scale acts as suppressor variable (Karol, (1994). (g)ns, p > 10. (*)p < .05. (**)p < .01. TABLE 3 Cross-Validated Regression of SDS From 16PF-5th Edition SDS Scale A-[R.sup.2] (R) (a) M SD Global factor score equations (**) Realistic (n = 229) .21 (.47) 17.6 5.1 Investigative (n = 230) .08 (.30) 21.7 3.1 Artistic (n = 230) .32 (.57) 24.5 6.7 Social (n = 229) .27 (.53) 36.2 4.6 Enterprising (n = 230) .17 (.42) 25.9 9.0 Conventional (n = 229) .05 (.24) 21.7 2.6 Primary factor score equations (**) Realistic (n = 230) .35 (.60) 17.1 6.5 Investigative (n = 229) .19 (.46) 21.4 4.9 Artistic (n = 230) .32 (.57) 24.3 6.8 Social (n = 230) .32 (.58) 36.7 9.0 Enterprising (n = 230) .20 (.46) 26.0 4.2 Conventional (n = 222) .03 (.19) 22.0 1.9 SDS Scale SEE 16PF Equation (b) Global factor score equations (**) Realistic (n = 229) 9.53 (-1.9*EX) + (-1.9AX) + (1.9*TM) + (3.2* IN) + (-0.8*SC) + 17.37 Investigative (n = 230) 10.19 (-1.9*EX) + (1.4*IN) + (-1.2*AX) + 32.36 Artistic (n = 230) 9.86 (-3.3*TM) + 40.20 Social (n = 229) 7.66 (1.4*EX) + (-1.7*TM) + (0.5*SC) + 33.14 Enterprising (n = 230) 8.18 (0.6*TM) + (2.2*IN) + 11.20 Conventional (n = 229) 10.61 (0.7*TM) + (0.9*SC) + + 13.71 Primary factor score equations (**) Realistic (n = 230) 8.65 (-3.0*1) + (-1.1*A) + (1.7*Q1) + 33.66 Investigative (n = 229) 9.55 (-1.8*A) + (-0.4*1) + (1.4*B) + (-0.7*Q4) + (1.2*Q1) + 23.64 Artistic (n = 230) 9.84 (2.1*1) + (1.2*M) + (1.4*Q1) - 3.92 Social (n = 230) 7.36 (2.0*A) + (0.8*H) + (1.0*Q1) + 13.85 Enterprising (n = 230) 8.06 (1.0*A) + (1.6*E) + (0.55*H) + (-0.8*1) + (0.7*N) + 9.87 Conventional (n = 222) 10.70 (-1.0*M) + 27.63 Note. See Table 1 Note. SEE = Standard Error of the Estimate. (a)A-[R.sup.2] is the adjusted [R.sup.2] for the equation with the multiple R in parentheses. (b)Letters represent scale names (Conn & Rieke, 1994). (**)p < .01.

Printer friendly Cite/link Email Feedback | |

Author: | Page, Betsy J. |
---|---|

Publication: | Career Development Quarterly |

Article Type: | Statistical Data Included |

Geographic Code: | 1USA |

Date: | Dec 1, 2001 |

Words: | 4786 |

Previous Article: | Gender and Ethnic Differences in Career Goal Attainment. (Articles). |

Next Article: | Coincidence, happenstance, serendipity, fate, or the hand of god: Case studies in synchronicity. (Article). |

Topics: |

## Reader Opinion