Vrijednost tekstualnih podataka u opservacijskim epidemiološkim istraživanjima [Value of textual data in observational epidemiological research]

Hercigonja-Szekeres, Mira (2010) Vrijednost tekstualnih podataka u opservacijskim epidemiološkim istraživanjima [Value of textual data in observational epidemiological research]. PhD thesis, Sveučilište u Zagrebu.

Preview

PDF Download (611kB) | Preview

Abstract

There are two established methodological approaches in present-day epidemiological research: quantitative and qualitative. Both approaches were used in this paper due to their complementary quality. Semi-structured interviews used with a refined analytical approach of statistical text analysis provide an opportunity to reach to the core and essence of the subject (personal estimate of their quality of life). The goal of this research was to compare the results of qualitative analysis of textual data acquired through general interview guided approach in epidemiological research with results acquired through analysis of data from SF-36 Health Survey. Additional goal was to compare the applicability of textual analysis done by computer programs and its contribution in gaining information on examinees in comparison with interpretative methods of textual analysis. Sample consist of 117 examinees aged from 23 to 82, more women (55,6%) than men, of various professions, most of them married (73,5%), more than 85% have secondary-school education, most of them are retired (59 %), more than half (52,2%) of them self-evaluated their health is the same as health of their peers, most of them (75%) live in Zagreb. This research used a Croatian edition of SF-36 as a validated and standardized questionnaire. Analysis of questionnaire SF-36 shows that the sample correlated to results gained through research based on the same questionnaire used on 1% of inhabitants in Croatian regions. The analysis of interview guided approach through interpretative techniques resulted in 12 clusters with different numbers of examinees. Every cluster was analyzed and described by examinees' characteristics, as well as by the Physical Component Summary (PCS) measure and the Mental Component Summary (MCS) measure from SF-36 Health Survey. The analysis of interview guided approach through computer program Dtm-Vic provided two kinds of results: results of lexicometrical analysis and results of correspondence and cluster analyses. Lexicometrical analysis of textual data provided 32365 words all together, 5381 or 16.4% are different words, which shows that examinees mostly use the same language in describing their health condition. 35 most frequently used words are: sam (od jesam) (= I am (from "to be")), ja (= I), mi (= to me), me (= me), imam (= I have), nemam (I don't have), ne_mogu (I can't), mogu (I can). That shows a high level of individual approach in heath evaluation which was insisted upon. Final results of correspondence and cluster analyses were 12 clusters of examinees with different numbers of examinees. They were analyzed and described by examinees' characteristics, as well as by the Physical Component Summary (PCS) measure and the Mental Component Summary (MCS) measure from SF-36 Health Survey. Clusters obtained by different techniques of analyses (interpretative and computer) cannot be mutually compared as they included different examinees. Comparison of questionnaire SF-36 analysis and textual data acquired through general interview guided approach shows great similarity in final results. However, textual data, i.e. their analysis, provide much more information, they are more detailed and show examinees' individuality. The paper shows that general interview guided approach is particularly suitable and useful in epidemiological researches because it provides plentiful information on population sample chosen for this research using the method of intensity sampling.

Abstract in Croatian

U epidemiološkim istraživanjima danas su se ustalila dva metodološka pristupa, kvantitativni i kvalitativni. U ovom su radu primjenjena oba postupka upravo zbog svoje komplementarnosti. Polustrukturirani intervjui uz primjenu rafiniranog analitičkog postupka statističke analize teksta otvaraju mogućnost ulaska u dubinu i bit ispitivanog problema (doživljaja kvalitete života). Ciljevi istraživanja bili su usporediti rezultate kvalitativne analize podataka dobivenih otvorenim ispitivanjima u epidemiološkim istraživanjima s rezultatima dobivenima analizom podataka iz Upitnika SF-36 te usporediti primjenjivost računalne analize tekstnih podataka i njezin doprinos saznanjima o ispitanicima s interpretativnom tehnikom. Uzorak čini 117 ispitanika u dobi od 23 do 82 godine, žena je nešto više (55,6%), različitih su zvanja, većinom su (73,5%) u braku, više od 85% ima završenu barem srednju školu, većinom su umirovljenici (59%), više od polovice (52,2%) samoprocjenjuje svoje zdravlje jednakim u odnosu na zdravlje svojih vršnjaka, većina (75%) ih je iz raznih dijelova Zagreba. Kao patentiran i validiran upitnik primijenjen je SF-36, hrvatsko izdanje. Rezultati analize upitnika SF-36 pokazuju da se rezultati statistički ne razlikuju od rezultata dobivenih ispitivanjem istim upitnikom oko 1% stanovnika svih regija Hrvatske. Analizom otvorenih ispitivanja interpretativnom tehnikom dobiveno je 12 klastera ispitanika različite veličine. Svaki je klaster analiziran i opisan značajkama ispitanika te profilima fizičkog i psihičkog zdravlja dobivenim Upitnikom SF-36. Računalna obrada tekstnih podataka programom Dtm-Vic kao rezultat leksikometrijske analize daje ukupni broj riječi 32365, od toga je 5381 ili 16,4% različitih riječi što pokazuje da se ispitanici uglavnom služe istim riječima u opisu svojeg zdravstvenog stanja, a učestalost pojedinih riječi pokazuje osobni pristup svojem zdravlju. Rezultat korespondencijske i klaster analize je 12 klastera ispitanika različite veličine. Ti su klasteri također opisani značajkama ispitanika te profilima fizičkog i psihičkog zdravlja dobivenim Upitnikom SF-36. Skupine (klasteri) dobivene različitim tehnikama (interpretativnom i računalnom) nisu međusobno usporedive zbog različitih ispitanika u njima. Usporedba analiza upitnika SF-36 i tekstnih podataka dobivenih razgovorom pokazuje veliku sličnost u krajnjim rezultatima. Međutim tekstni podaci, odnosno rezultati njihovih analiza donose mnogo više informacija, mnogo su bolje nijansirani, čuvaju individualnost ispitanika. Rad pokazuje da su tematski ciljane ankete kroz razgovor osobito primjenljive i korisne u epidemiološkim istraživanjima zbog bogatstva informacija na uzorku populacije koji je u ovom istraživanju bio izabran metodom intensity sampling.

Actions (login required)

View Item

Downloads