Skip to main content Skip to main navigation menu Skip to site footer
Research Articles
Published: 2020-08-21

Features of statistical analysis of medical-psychological research

Bukovinian State Medical University
Bukovinian State Medical University
statistical analysis psychology measurement scale

Abstract

The purpose of the study: to consider the features of statistical analysis in psychological and medical-psychological research.

Results. One of the features of psychological research is the process of measuring the results of the experiment. Unlike medical research, which typically uses physical units of measurement, most variables in psychological research are not unambiguous or easy to measure. To describe the procedures of psychological measurement in psychological research four types of measurement scales are used: nominative, ordinal, interval, and scale of equal relations. Statistical analysis of the results of medical and psychological research depends on the type of scale in which the studied trait was measured.

The nominative scale is a scale that classifies by name. Conjugacy tables are used to describe and analyze nominative scales. An ordinal scale is a scale that classifies on the principle of "more or less". Statistical analysis of ordinal scales is performed using non-parametric criteria. The interval scale is a scale according to which each of the possible values ​​of the feature is at the same distance from the other value. The scale of equal relations has all the properties of nominative, ordinal, and interval scales. To analyze the results of the study, which were measured in interval scales or in scales of equal relations, one uses parametric or nonparametric criteria depending on the distribution of a random variable.

Conclusions. The choice of method of statistical analysis of the results of psychological research depends on the type of scale in which the data were measured.

Background

In the last two decades in world medicine, there has been an increase in the use of the interdisciplinary approach in research and multidisciplinarity as a principle of providing not only medical care but also ensuring the quality of life of the patient and their physical and socio-psychological well-being observed [1,2]. A specific place in this context is held by medical and social studies. First of all medical and psychological researches; the psychology of health, the psychosomatic approach is now gaining special importance [3,4]. According to its methods, psychology, as a science that has been formed relatively recently, differs significantly from medicine and other sciences. The main methods of psychology are observation, questioning, testing, experiment [5].

An important difference in psychological research is the process of measuring the results of a psychological experiment. Unlike physical units, for which measurement is a numerical estimate and expression of one quantity in relation to another, in psychometry measurement is the assignment of numbers to objects or events according to some rule [6]. In other words, measurement in the procedure of psychological experiment is considered as a method of registering the state of the object of study and, accordingly, changes in this state in response to the experimental effect [7]. Many variables studied by psychologists are easy to measure (age, height, weight). However, most variables in psychological research are not so unambiguous or easy to measure. We cannot, for example, accurately assess the level of intelligence of people, measure their self-esteem [8]. In addition, many psychological variables can be identified in different ways. For example, stress can be determined by a scale of stress, a scale of perceived stress, the concentration of cortisol in saliva, the number of stressful life events that the patient has recently experienced, or even a self-created indicator [8].

In psychology, there are three main procedures of psychological measurement, which are based on the object of measurement: the measurement of subjects, measurement of stimuli, joint measurement of stimuli, and subjects [7]. To describe the procedures of psychological measurement in psychological research using 4 types of measurement scales, which were proposed by S. Stevens [6].

Nominal level

The nominal level is a scale that classifies by name. The names are not quantified, they only distinguish one object of study from another. The simplest nominative scales are dichotomous scales that have only two categories (for example, gender, dominant hand, the presence of siblings) [9]. More complex nominative scales consist of three or more cells. These can be, for example, classified by ethnicity, marital status, eye color. An essential feature of nominal scales is that they do not provide any ordering of answers. For example, by classifying people by their favorite color, the researcher cannot put a green "ahead" of blue. The answers are simply classified. Thus, the nominal scales reflect the lowest level of measurement [8]. Operating on nominative scales, the researcher has only one numerical characteristic - the number of observations.

For statistical analysis of nominal level, conjugation tables are used, which indicate the number of persons in each experimental group who have or do not have this feature. The research results are presented in the form of quantity (%). Pearson's test is used to compare the frequency of the trait in the two experimental groups, provided that the samples are independent and in all cells, the number is greater than 10. To compare small independent samples, Fisher's exact criterion is used. If it is necessary to compare dependent samples, the McNemar test is used. To assess the relative risk in case-control studies, the odds ratio is used, and to analyze prospective studies, the relative risk is used. The analysis of conjugacy tables is described in more detail in the work of M. Ivanchuk et al. [10].

Let’s consider an example of data presented at the nominal level. Hereinafter, examples of statistical calculations are based on the examination of patients with cardiac arrhythmias who were in inpatient or outpatient treatment at the Chernivtsi Regional Clinical Cardiology Center. The sex distribution of patients with different nosologies (extrasystoles of high grades, atrial fibrillation, paroxysmal tachycardia, conduction disorders, including sinoatrial, atrioventricular block, and sinus weakness syndrome) is studied. Since it is not possible to order the nosology data from the "largest" to the "smallest", for statistical analysis of the data we consider the data like those presented in the nominative level.

Data are presented in Table 1 as n (%).

Men Women
High-grade extrasystoles 84 (52.5) 76 (47.5)
Atrial fibrillation 44 (55.0) 36 (45.0)
Paroxysmal tachycardia 26 (53.1) 23 (46.9)
Violation of conductivity 12 (52.2) 11 (47.8)
Table 1. Table 1 Gender distribution of patients with different nosologies.

To compare the distribution of patients by sex in groups relative to nosologies, we use Pearson's criterion, because the samples are independent, groups to compare more than two, in each cell of the table, the number is greater than 10. We obtain the value of the criterion. Therefore, the distribution of patients by sex in groups with different nosologies does not differ significantly.

Ordinal scale

Ordinal scale - a scale that classifies on the principle of "more or less". There must be at least three classes in the ordinal scale. In contrast to the nominal level, in the ordinal scale, the elements from largest to smallest can be ranked. However, unlike the interval scale and the ratio scale, the difference between the two levels cannot be considered the same as the difference between the other two levels. For example, if the satisfaction scale is considered, the same difference between "fully satisfied" and "partially satisfied" and between "partially satisfied" and "partially dissatisfied" cannot be considered [8].

Quantities and percentages are used to describe the data in the ordinal scale similarly to the nominal level. Statistical analysis of ordinal scales is performed using nonparametric methods. The Mann-Whitney test or Rosenbaum test is used to compare independent samples, and the Wilcoxon test or the sign test is used to compare dependent samples. To compare three or more levels of a particular trait using the Kruskal-Wallis test. Spearman's rank correlation is used to determine the consistency of changes in traits.

Let’s consider an example of data presented on an ordinal scale. The distribution of patients by sex according to the level of alexithymia is studied. According to the results of the study using the Toronto scale of alexithymia, three levels are determined - the absence of alexithymia, trend (intermediate value), and the presence of alexithymia. Since there are three classes that can be ranked from the smallest (no alexithymia) to the largest (existing alexithymia), statistical analysis of data is performed as an analysis of data presented on an ordinal scale.

The results of the study are shown in table 2 as n (%).

Men Women
Absence 48 (49.5) 49 (50.5)
Trend 90 (63.4) 52 (36.6)
Alexithymia 68 (60.7) 44 (39.3)
Table 2. Table 2 Distribution of patients by sex according to the level of alexithymia

To compare the groups, we rank for the levels of alexithymia: absence - 1, trend - 2, alexithymia - 3 and use the Mann-Whitney test. We conclude that there is no difference between the levels of alexithymia in patients of different sexes (p> 0.05).

Interval scale

The interval scale is a scale for which each of the possible values ​​of the features is at the same distance from the other values ​​[9]. The disadvantage of interval scales is that they do not have a true zero point, even if one of the scaled values ​​is called "zero". An example of an interval scale in psychology is the level of intelligence (IQ). Even if it is technically possible to obtain a score of 0 at the IQ level, such a score will not indicate a complete absence of IQ. Moreover, if a person receives an IQ of 140, it does not mean that his level of intelligence is twice as high as the level of intelligence of a person with an IQ of 70. However, the difference between IQ 80 and 100 is the same as the difference between IQ 120 and 140 [8].

The description of the data presented in the interval scale is presented using the mean value and its error or standard deviation, if the hypothesis of the normality of the distribution of a random variable is accepted, or using the median and interquartile range otherwise. More detailed methods for checking the distribution for normality are described in our previous work [11]. Statistical analysis of interval scales is performed using parametric or nonparametric criteria depending on the distribution of a random variable. If the hypothesis of a normal distribution of samples is accepted, then Student's criterion is used for their comparison (even for dependent samples and odd for independent ones). If it is necessary to compare three or more levels of a certain feature, perform Fisher's analysis of variance. Pearson's correlation analysis is used to determine the consistency of trait changes. If the distribution of at least one of the studied samples differs from normal, use nonparametric methods similar to the ordinal scales.

Let us consider an example of data presented on an interval scale. Similar to the previous example, we study the distribution of patients by sex according to the level of alexithymia. However, in this case, we do not take the level of alexithymia (absence-trend-alexithymia) for the study, but the scores obtained on the Toronto alexithymia scale. Since we cannot say that, for example, a patient with a score of 80 has twice the level of alexithymia than a patient with a score of 40, we assume that the data are presented on an interval scale.

The results of the study are shown in table 3 in the form of Mm because both samples are distributed normally.

Alexithymia
Men (n=206) 69.010.67
Women (n=145) 67.481.10
Table 3. Table 3 The level of alexithymia in patients of different sexes

To compare the samples, we use the odd Student's criterion. We conclude that there is no difference in the level of alexithymia in patients of different sexes (p = 0.219).

Ratio scale

The scale of equal relations has all the properties of nominal, ordinal, and interval scales. Like the nominal level, it provides a name or category for each object (figure serves as a score). As with the ordinal scale, objects are ordered (in terms of ordering numbers). As in the interval scale, the same difference in the two places on the scale has the same value. The main difference in the scale of equal relations is the presence of a real zero point [8]. A classic example of the difference between an interval scale and a scale of equal relations is the temperature scale. The Kelvin temperature is absolutely zero, and we can say that the temperature of 100°K is twice the temperature of 50°K. The Celsius temperature has no absolute zero, and therefore it cannot be said that a temperature of 100°C is twice as high as a temperature of 50°C. An example of a scale of equal relations in psychology is the scale of thresholds of absolute sensitivity. It should be noted that in psychology, the scale of equal relations is rare because the possibilities of the human psyche are so great that it is difficult to find a measurable psychological variable that has an absolute zero [9].

Statistical analysis of scales of equal relations is carried out similarly to the analysis of interval scales.

Consider an example of data presented in the scale of equal relations. The distribution of nosologies by age is studied. Age is one of the few variables studied in psychology, which is presented on the scale of equal relations because it has an absolute zero, and a person aged 50 can be said to be twice as old as a person aged 25.

The results of the study are shown in table 4 in the form of Me (IQR) because the distribution of the samples is different from normal.

Age
High gradation extrasystole (n=123) 49 (24)
Atrial fibrillation (n=120) 65.5 (11.25)
Paroxysmal tachycardia (n=49) 44 (23)
Conduction disturbances (n=22) 62.5 (15)
Table 4. Table 4 Age distribution of patients by nosology

To compare the age in the four groups, we use the nonparametric Kruskal-Wallis test. We conclude that there is a difference in group means (p <0.001).

The differences between the scales and statistical methods used for their analysis are summarized in table 5.

Scale Category scores Ranking procedure Regular intervals True zero Descriptive statistics Statistical analysis
Normal distribution The distribution is different from normal Normal distribution The distribution is different from normal
Dependent samples Independent samples Dependent samples Independent samples
Nominal + n (%) McNemar test relative risks Pearson's χ2 (n≥10); Fisher's exact criterion (n <10); odds ratio
Ordinal + + n (%) Mann-Whitney criterion; Rosenbaum's criterion Wilcoxon's criterion; Criterion of signs; Spearman's rank correlation; Kruskal-Wallis criterion
Interval + + + M±m M±SD Me (IQR) Student's paired criterion Student's unpaired criterion; Pearson correlation; Fisher's analysis of variance Mann-Whitney criterion; Rosenbaum's criterion Wilcoxon's criterion; criterion of signs; Spearman's rank correlation; Kruskal-Wallis criterion
Ratio + + + + M±mM±SD Me (IQR) Student's paired criterion Student's unpaired criterion; Pearson correlation; Fisher's analysis of variance Mann-Whitney criterion; Rosenbaum's criterion Wilcoxon's criterion; criterion of signs; Spearman's rank correlation; Kruskal-Wallis criterion
Table 5. Table 5 Measurement scales and their statistical analysis

Conclusion

The choice of statistical analysis method of the results of psychological research depends on the type of scale in which the data were measured. Correct application of statistical analysis is the key to obtaining reliable results of medical and psychological research.

References

  1. Gagne T, Lapalme J, McQueen DV Multidisciplinarity in health promotion: a bibliometric analysis of current research Health Promotion International. 2017:1–12. https://doi.org/ 10.1093/heapro/dax002 https://pubmed.ncbi.nlm.nih.gov/28334790/
  2. Barthel R, Seidl R Interdisciplinary Collaboration between Natural and Social Sciences – Status and Trends Exemplified in Groundwater Research. PLoS ONE. 2017; 12(1): e0170754. https://doi.org/10.1371/journal.pone.0170754 https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0170754
  3. Johnson BT, Acabchuk RL What Are the Keys to a Longer, Happier Life? Answers from Five Decades of Health Psychology Research Soc Sci Med. 2018 Jan; 196: 218–226. https://doi.org/ 10.1016/j.socscimed.2017.11.001 https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6894515/
  4. Chaban OS, Khaustova OO. Psychosomatic comorbidity and quality of life in elderly patients. NeuroNEWS. 2016; 1(2):8-12. URL: https://neuronews.com.ua/uploads/issues/2016/2-1/803578556.pdf
  5. Turynina OL. Metodolohiia ta metody psykholohichnoho doslidzhennia. Kyiv: Personal; 2018. 206. http://maup.com.ua/assets/files/lib/book/metod_psih_dosl.pdf
  6. Styvens SS, editor. Matematyka, yzmerenye y psykhofyzyka. Eksperymentalnaia psykholohyia. М; 1960: 19–89.
  7. Odynochkyna EV. Uchebno-metodycheskyi kompleks po uchebnoi dystsyplyne «Metody psykholohycheskoho yssledovanyia». Gomel; 2019. 152. https://elib.bsu.by/handle/123456789/213940
  8. Jhangiani RS., Chiang IA., Cuttler C., Leighton D. Research Methods in Psychology, 4th Edition. Kwantlen Polytechnic University. 2019. 432 p. https://kpu.pressbooks.pub/psychmethods4e/
  9. Sydorenko EV. Metody matematycheskoi obrabotky v psykholohyy, SPb: Rech; 2001. 350 с. http://umo.edu.ua/images/content/aspirantura/zabezp_discipl/sidorenko.pdf
  10. Ivanchuk MA., Malyk IV., Knihnitska TV., Lukashiv TO. Statystychnyi analiz vidnosnykh velychyn v medytsyni. Klinichna ta eksperymentalna patolohiia. 2019;4 (70);109-114 http://cep.bsmu.edu.ua/article/view/1727-4338.XVIII.4.70.2019.18 https://doi.org/10.24061/1727-4338.XVIII.4.70.2019.18
  11. Ivanchuk MA., Ivanchuk PR. Normalnyi zakon rozpodilu v medychnykh doslidzhenniakh. Medychna informatyka ta inzheneriia. 2013 (1):48-52 https://doi.org/10.11603/mie.1996-1960.2013.1.419

How to Cite

1.
Ivanchuk М, Polishchuk О. Features of statistical analysis of medical-psychological research. PMGP [Internet]. 2020 Aug. 21 [cited 2024 Mar. 29];5(3):e0504255. Available from: https://ojsdemo.e-medjournal.com/index.php/psp/article/view/255