Suicide Research and Non-Rejection of the Null Hypothesis

Abstract

Research results on suicide have been scanty and inconsistent. Because of the peculiar conditions (paucity of Ss, variability of research materials, and selection of treatment conditions) found in suicide research, it is felt that the Type 11 error (failing to reject the null hypothesis when it is false) is probably committed too often. It is suggested chat less stringent significance levels should be accepted, as this might facilitate identification of new variables.There has been an increasing research interesc in s~~i c i d e in the last decade and a half, with many investigations using suicide notes, personality and projective inventory measures, as well as problem-solving techniques. However, adequate description, identification and prediction of suicidal ideation or behavior is, as yet, not feasible because of sparse and inconsistent research findings.There are several reasons for these difficulties, chief among them being the scarcity of cases. Stengel and Cook ( 3 6 ) have likened the problem facing the suicide investigator co that of the pathologist trying to establish the cause of death without access to the corpse. The paucity o f cases raises some very difficult problems. T o overcome this deficiency several methods have been used: analysis of residuals such as suicide notes, diaries, psychological tests administered before the death, social histories, and the memories of friends and relatives, and smdy of live Ss as substitutes for suicidal persons. None of these seems adequate from a methodological standpoint. Rosen ( 2 7 ) , commencing on the observation that suicide is a highly infrequent event, has expressed doubts about the feasibility of predicting suicidal behaviors from current assessment techniques.T h e authors reviewed the available literamre on suicide in which comparative statistics were reported (1, 8 ) . The authors suspect that there have been more such studies, but they have not been made public because of non-significant results. T h e sample size in these scudies range from 3 ( 3 2 ) to 948 ( 3 3 ) . The largest N s are found in studies of suicide notes, for usually chese have accumulated over several decades, frequently in the files of medical examiners and coroners. The next largest N s are for studies using MMPI test protocols. Again, for this popular test protocols accumulate in the files of neuropsychiatric hospitals and clinics. The very small N s occur in studies dealing with projective tests or when some novel approach is used. Projective data are quite limited and swapping occurs fairly often ( 4 , 7, 3 2 ) . In other research Ns are 3 to 30 depending upon whether committed or attempted suicide is studied. As small N s diminish the probability that the nil11 hypothesis will be re-

Attempts to predict suicidal risk using psychological tests.

“…A final point made by Neuringer is that in view of the difficulty evidenced by research workers in finding variables that differentiate suicidal from nonsuicidal persons, a more liberal level of significance than 5% be used so that valid hypotheses will not be rejected (Neuringer & Kolstoe, 1964). In one way, this makes sense.…”