Journal of Psychosomatic Research, Vol. 20, pp. 201 to 205. Pergamon Pm,

1976. Printed in Great Britain

SAMPLE SIZE CHARACTERISTICS AND SCORES ON THE SOCIAL READJUSTMENT RATING SCALE LINAS

A.

BIELIAUSKAS*

and

DEBRA

A. S’rnuciARt

THE Social Readjustment Rating Scale (SRRS) has been developed by Holmes and Rahe [l] as a measure of the relationship between stressing life events and evidence of physical or psychological maladjustment. Significant relationships have been reported between SRRS scores and signs and symptoms of everyday life [2], probability of illness [3-61, and incidence of psychiatric hospitalization [7-91. However, questions have been raised concerning the predictive utility of the SRRS. Rubin et al. [lo] found no significant differences in occurrence of life-stress events between hospitalized and non-hospitalized crewmen aboard a navy ship. Aponte and Miller [l l] found little relationship between life-stress events and current psychiatric patient status. Bieliauskas and Webb [12] reported significant relationships between scores on a standard SRRS and a college-modified version of the SRRS and incidence of aidseeking for physical or mental health reasons in college students. The magnitude of these relationships, however, was found to be unimpressive and no more than three percent of subject population variance was explained by SRRS scores. This finding further questioned the practical utility of the SRRS as a predictive instrument. The SRRS has generally been employed as a group measure with differing subject populations and sample sizes. If the demonstrated relationship between SRRS scores and incidence of maladjustment is of significant but limited magnitude [12], sample size characteristics may be affecting the significance of reported results. It was the main purpose of this study to examine the significance and magnitude of the relationship between scores on a college-modified version of the SRRS [12] and aid-seeking in different size college student populations. In addition, since the frequency of occurrence of a specific life-stress event may be related to SRRS validity [12], SRRS scores were evaluated both in terms of standard life change units [I] yielding standard weight scoring (SW), and frequency x life change unit (FW). Finally, though generally good test-retest reliability coefficients [13, 141and reliable item ranking over differing subject populations [15, 161 have been demonstrated for the standard SRRS, the test itself is frequently modified across studies. Though these modifications are generally minor, e.g. [3, 171, earlier reliability estimates might not continue to apply. This study examined the test-retest reliability of the collegemodified version of the SRRS and the reliability of item rankings between test administrations. METHOD

Subjects Undergraduate psychology students who completed a paper and pencil test booklet were subjects in the three groups used to examine sample size characteristics. Group A consisted of 116 females and 137 males; Group B, 71 females and 51 males; Group C, 30 females and 23 males. *Dept. of Clinical Psychology, Shands Hospital, Univ. of Florida, Gainesville, FL 32601, U.S.A. t Dept. of Int. Communication, Ohio University, Athens, OH 45701, U.S.A. 201

202

LINAS A. BIELIAUSKASand DEBRA A. STRUGAR

Test-retest reliability and inter-administration item ranking was examined in a fourth group of subjects consisting of 10 female and 13 male undergraduate communication students. Materials A test booklet containing one item to indicate sex, one item to indicate aid-seeking, and 46 items of the college-modified ERRS [12] was administered to all subjects. Each of the SRRS items included 5 choices to indicate the frequency with which each life-stress event had occurred in the six months preceding test administration. Procedures The test booklet was completed in one sitting for each group administration. Verbal and written instructions were identical to those described elsewhere [12]. The responses on all answer sheets for Groups A, B, and C were then transferred to computer cards and scored both for single SRRS weights per item (SW) and for item weights multiplied times the frequency of occurrence for each item event (FW). To assess SRRS reliability, the test was re-administered to the final group of subjects two weeks after the first test administration. All computations for this group were performed by hand. RESULTS Sample size characteristics. Sample sizes were smaller by a factor slightly more than half in the succeeding Groups A, B, and C. Group B was 0.48 the size of Group A; Group C was 0.42 the size of Group B. Means and standard deviations of SRRS scores for subjects in each group are presented in Table 1. In all groups, the percentage of subjects seeking aid ranged between 37 % (Group A) and 32 % (Group C). TABLE 1 .-GROUP

SIZEAND TESTSCORE CHARACTERISTICS Scoring method

Group A N Mean score on SRRS Standard deviation Mean score on SRRS Standard deviation Group B N Mean score on SRRS Standard deviation Mean score on SRRS Standard deviation Group C N Mean score on SRRS Standard deviation Mean score on SRRS Standard deviation

SW FW

SW FW

SW Fw

Aid-seeking subjects

Non aid-seeking subjects

94 (37 %) 381.49 304.52 759.52 559.06

159 (63 %) 286.89 206.29 562.37 476.22

43 (35 %) 277.74 172.13 523.02 409.24

79 (65 %) 229.54 145.66 390.63 324.50

I7 (32%) 286.06 224.24 523.82 480.55

36 (68 %) 263.92 163.31 447.14 369.06

Computation of the Dunn Multiple Comparison statistic [18] demonstrated that SRRS scores were significantly higher for aid-seeking subjects than for non aid-seekers in Group A (JJ < 0.01, df = 251), but this difference was not significant from zero for Group B (p > 0.05, df = 120) or Group C (p > O-05, df = 51). It should also be noted that the absolute value of SRRS scores was greater in Group A for both aid-seekers and non aid-seekers than for aid-seekers in the other two groups. Point biserial correlation coefficients were then computed between the results of both SW and FW SRRS scores and incidence of subject aid seeking during the six months prior to testing. Both scoring methods correlated significantly with aid-seeking in the largest group (A) at the p < 0.01 level. However, as can be seen in Table 2, the significance of this relationship decreased to the p < 0.05 level in the next largest group (B), and was no longer different from zero (p > 0.10, df = 51) for Group C.

Sample size characteristics TABLE

2.--POINT

BISERIAL

and scores on the Social Readjustment

CORRELATION, GROUP

Group A

Scoring method

N = 253

SW FW SW FW SW FW

N=

122

C

N=

53

tp < 0.05.

OF VARIANCE,

SCORES

Sample size

B

*p < 0.01.

ANALYSIS

SRRS

AND

AND

OMEGA

Rating Scale SQUARE

RESULTS

203 BETWEEN

AID-SEEKING

r 0.18* @23* 0.1st 0.18t 0.06 0.09

F-value 8.65* 8*91* 2.681 3.84$ 0.17 0.41

Omega square value 0.03 0.03 0.01 0.02 0.00 0.00

$p < 0.10.

A oneway analysis of variance was similarly performed for both SRRS scoring conditions to further assess the relationship between aid-seeking and test scores in the different groups. For both SW and FW scoring conditions, the difference between subjects receiving aid and those not was highly significant in Group A 0, < 0.01, df = l/252). In Group B this difference was significant only at the p < 0.10 level. The strength of this relationship was not significantly different from zero in Group C (see Table 2). The proportion of variance accounted for by SRRS score differences, as indicated by the Omega Square statistic 1181,showed a similar decline from a value of 0.03 in Group A, to 0.01-0*02 in Group B, and 0.00 in Group C (see Table 2). No significant differences in scoring efficiency were noted between SW and FW scoring methods. ReliabiIity characteristics. A Pearson Product Moment correlation coefficient was computed between SRRS scores achieved two weeks apart by subjects in the fourth group. Good test-retest reliability was demonstrated for the college-modified SRRS (r = 0.82, df= 21, p < 0901). To assure the reliability of item ranking within the college-modified SRRS, Kendall’s Coefficient of Concordance [19] was computed for the frequency of individual item choice by subjects in the two separate test administrations. The frequency ranking of items selected was highly reliable between test administrations (W = 0.91, N = 46, p < 0.01). As a tial computation, a point biserial correlation coefficient was performed, using the SW scoring method, between SRRS scores and incidence of aid-seeking in this final small (N = 23) group. The trends evidenced above, between SRRS score significance and group size, continued (1st administration: r = -0.03, df = 21, p > 0.10; 2nd administration: r = 0.051, df = 21, p > 0.10). DISCUSSION

It becomes aid-seeking population

clear that

the power

subjects in a college sample being measured

of the SRRS population decreases.

to discriminate

aid-seeking

steadily decreases as the This remains true regardless

from

non

size of the of whether

SW or FW scoring method is used to compute test scores. In light of sampling theory, which predicts that larger samples are more reliable than smaller ones and that larger samples take into account a greater proportion of extreme measures (such as SRRS scores) in the population to which the events are being generalized [19], the current findings are not unexpected. The larger standard deviations (and consequent variance of scores) between Group A and the other groups would seem to confirm this observation (see Table 1). Group C, however, has a slightly larger standard deviation than does Group B. While objective reasons for this result are unclear, these values are consistent with an increase in the standard error of the mean as sample sizes decrease [19]. This would suggest that with samples approximately the size of Group C, bias in measurement of SRRS score population distribution is likely to occur. Though these results suggest SRRS score measurements generally conform to sampling theory expectations, they augur poorly for use of the SRRS as a utilitarian predictive instrument of aid-seeking probability. Differences in mean SRRS scores between aid-seeking and non aid-seeking subjects in the different groups (see Table 1) reveal another interesting finding. The mean SRRS the

204

LINAS A. BIELIAUSKASand DEBRA A. SYRUGAR

score for non aid-seekers in Group A is larger than the mean SRRS scores for aidseekers in Groups B and C. This occurs in spite of the fact that the non aid-seeker mean SRRS score in Group A is significantly lower than the aid-seeking score in the same group. These values suggest that either generally larger absolute value SRRS scores are obtained in larger sample sizes or SRRS score discrimination power between aid-seeking and non aid-seeking subjects is more potent in populations which experience a greater number of life change events. While the former interpretation is certainly consistent with the generally larger range of scores recorded from larger samples, the latter certainly warrants further investigation in relation to the kinds of populations to which sample results will be generalized. One final characteristic of the data presented thus far is that between 32 and 37 % of subjects, or approximately one in three, randomly sampled from a normal college population, will have sought aid for physical or mental health reasons in the previous six months. Though such an incidence merits consideration in determining sizes of samples for future study of aid-seeking and life event relationships in normal college students, it is also a statistic worthy of further research concerning student academic difficulties and prevention program effectiveness in physical and mental health areas. Use of the college-modified version of the SRRS is justified in terms of test-retest reliability and item selection reliability, both of which are consistent with values reported for the standard SRRS. The general results of this study suggest that while the SRRS may accurately record the occurrence of possibly stressing life events, situational or individual characteristic variables which modulate the actual stress response of any given subject to those events are not measured. One event may cause stress in one individual while not affecting another. Such subject differences would be randomly distributed in large samples and cancel each other out. In smaller samples, however, they would assume far greater importance. We are currently investigating methods of measuring such variables in our laboratories. For the present, the uncertainty concerning the utilitarian efficiency of the SRRS remains. SUMMARY The relationship between Social Readjustment Rating Scale (ERRS) scores and incidence of aid-seeking is examined in college population samples of different sizes. Reliability characteristics of a college-modified version of the SRRS are also studied. The results generally suggest that the power of the SRRS to discriminate aid-seeking from non aid-seeking college students decreases as the size of the sample being tested decreases. Possible reasons for the current findings as well as their implications are discussed. REFERENCES

11,213 1. HOLMEST. H. and RAHE R. H. The social readjustment rating scale. J. Psychosom. Res. (1967). 2. HOLMEST. S. and HOLMEST. H. Short-term intrusions into the life style routine. J. Psychosom. Res. 14,121 (1970). 3. RAHE R. H.. MCLEAN J. D. and ARTHUR R. J. A longitudinal study of life-change and illness patterns. J. ksychosom. Res. 10,355 (1967). 4. RAHE R. H.. MAHAN J. L. and ARTHUR R. J. Prediction of near-future health change from subjects’ pm&ding life changes. J. Psychosom. Res. 14, 401 (1970). 5. CLINED. W. and CHOSYJ. J. A prospective study of life changes and subsequent health changes. Archs Gen. Psych&d. 27, 51 (1972). 6. RAHE R. H. Subjects’ recent life changes and their near-future illness reports. Ann. Clirr. Res. 4, 150 (1972).

Sample size characteristics

and scores on the Social Readjustment

Rating Scale

205

7. PAYKELE. S., MYERSJ. K., DIENELTM. N., KLERMANG. L., LINDERTHAL.I. J. and PEPPERM. P. Life events and depression. Arch. Gen. Psychiut. 21, 753 (1969). 8. UI-ILEN~U~HE. H. and PAYKELE. S. Symptom intensity and life events. Arch. Gen. Psychiut. 28,473 (1973). 9. DEKKERD. J. and WEBB J. T. Relationships of the social readjustment rating scale to psychiatric patient status, anxiety and social desirability. J. Psychosom. Res. 18, 125 (1974). 10. RIJBIN R. T., GUNDERSONE. K. E. and ARTHURR. J. Life stress and illness patterns in the U.S. navy-V. Prior life change and illness onset in a battleship’s crew. J. Psychosom. Res. 15, 89 (1971). 11. APONTE J. F. and M~LER F. T. Stress-related social events and psychological impairment. J. Cl&. Psychol. 28, 455 (1972). 12. BIELIAUSKASL. A. and WEBB J. T. The social readjustment rating scale; validity in a college population. J. Psychosom. Res. 18, 115 (1974). 13. CASEYR. L., MASUDAM. and HOLMEST. H. Quantitative study of recall of life events. J. Psychosom. Res. 11, 239 (1967). 14. THURL~W H. J. Illness in relation to life situation and sick-role tendency. J. Psychosom. Res. 15,73 (1971). 15. RUCH L. 0. and HOLMEST. H. Scaling of life change: comparison of direct and indirect methods. J. Psychosom. Res. 15, 221 (1971). 16. KOMAROFPA. L., MASUDAM. and HOLMEST. H. The social readjustment rating scale: a comparative study of Negro, Mexican, and white Americans. J. Psychosom. Res. 12, 121 (1968). 17. MASUDA M. and HOLMEST. H. Magnitude estimations of social readjustments. .I. Psychosom. Res. 11,219 (1967).

18. KIRK R. E. Experimental Design: Procedures for the Behavioral Sciences. Brooks-Cole, Belmont, California (1968). 19. DOWNIE N. M. and HEATH R. W. Basic Statistical Methods (2nd Ed.). Harper & Row, New York (1965).

Sample size characteristics and scores on the Social Readjustment Rating Scale.

Journal of Psychosomatic Research, Vol. 20, pp. 201 to 205. Pergamon Pm, 1976. Printed in Great Britain SAMPLE SIZE CHARACTERISTICS AND SCORES ON TH...
407KB Sizes 0 Downloads 0 Views