Thesis (MSc)--Stellenbosch University, 2004. / ENGLISH ABSTRACT: Any branch in science experiences measurement error to some extent. This maybe due to
conditions under which measurements are taken, which may include the subject, the
observer, the measurement instrument, and data collection method. The inexactness
(error) can be reduced to some extent through the study design, but at some level further
reduction becomes difficult or impractical. It then becomes important to determine or
evaluate the magnitude of measurement error and perhaps evaluate its effect on the
investigated relationships. All this is particularly true for blood pressure measurement.
The gold standard for measunng blood pressure (BP) is a 24-hour ambulatory
measurement. However, this technology is not available in Primary Care Clinics in South
Africa and a set of three mercury-based BP measurements is the norm for a clinic visit.
The quality of the standard combination of the repeated measurements can be improved
by modelling the measurement error of each of the diastolic and systolic measurements
and determining optimal weights for the combination of measurements, which will give a
better estimate of the patient's true BP. The optimal weights can be determined through
the method of structural equations modelling (SEM) which allows a richer model than the
standard repeated measures ANOVA. They are less restrictive and give more detail than
the traditional approaches.
Structural equations modelling which is a special case of covariance structure modelling
has proven to be useful in social sciences over the years. Their appeal stem from the fact
that they includes multiple regression and factor analysis as special cases. Multi-type
multi-time (MTMT) models are a specific type of structural equations models that suit
the modelling of BP measurements. These designs (MTMT models) constitute a variant
of repeated measurement designs and are based on Campbell and Fiske's (1959)
suggestion that the quality of methods (time in our case) can be determined by comparing
them with other methods in order to reveal both the systematic and random errors. MTMT models also showed superiority over other data analysis methods because of their
accommodation of the theory of BP. In particular they proved to be a strong alternative to
be considered for the analysis of BP measurement whenever repeated measures are
available even when such measures do not constitute equivalent replicates. This thesis
focuses on SEM and its application to BP studies conducted in a community survey of
Mamre and the Mitchells Plain hypertensive clinic population. / AFRIKAANSE OPSOMMING: Elke vertakking van die wetenskap is tot 'n minder of meerdere mate onderhewig aan
metingsfout. Dit is die gevolg van die omstandighede waaronder metings gemaak word
soos die eenheid wat gemeet word, die waarnemer, die meetinstrument en die data
versamelingsmetode. Die metingsfout kan verminder word deur die studie ontwerp maar
op 'n sekere punt is verdere verbetering in presisie moeilik en onprakties. Dit is dan
belangrik om die omvang ven die metingsfout te bepaal en om die effek hiervan op
verwantskappe te ondersoek. Hierdie aspekte is veral waar vir die meting van bloeddruk
by die mens.
Die goue standaard vir die meet van bloeddruk is 'n 24-uur deurlopenee meting. Hierdie
tegnologie is egter nie in primêre gesondheidsklinieke in Suid-Afrika beskikbaar nie en
'n stel van drie kwik gebasseerde bloedrukmetings is die norm by 'n kliniek besoek. Die
kwaliteit van die standard kombinasie van die herhaalde metings kan verbeter word deur
die modellering van die metingsfout van diastoliese en sistoliese bloeddruk metings. Die
bepaling van optimale gewigte vir die lineêre kombinasie van die metings lei tot 'n beter
skatting van die pasiënt se ware bloedruk. Die gewigte kan berekening word met die
metode van strukturele vergelykings modellering (SVM) wat 'n ryker klas van modelle
bied as die standaard herhaalde metings analise van variansie modelle. Dié model het
minder beperkings en gee dus meer informasie as die tradisionele benaderings.
Strukurele vergelykings modellering wat 'n spesial geval van kovariansie strukturele
modellering is, is oor die jare nuttig aangewend in die sosiale wetenskap. Die aanhang is
die gevolg van die feit dat meervoudige lineêre regressie en faktor analise ook spesiale
gevalle van die metode is. Meervoudige-tipe meervoudige-tyd (MTMT) modelle is 'n
spesifieke strukturele vergelykings model wat die modellering van bloedruk pas. Hierdie
tipe model is 'n variant van die herhaalde metings ontwerp en is gebaseer op Campbell en
Fiske (1959) se voorstel dat die kwaliteit van verskillende metodes bepaal kan word deur
dit met ander metodes te vergelyk om sodoende sistematiese en stogastiese foute te
onderskei. Die MTMT model pas ook goed in by die onderliggende fisiologies aspekte van bloedruk en die meting daarvan. Dit is dus 'n goeie alternatief vir studies waar die
herhaalde metings nie ekwivalente replikate is nie.
Hierdie tesis fokus op die strukturele vergelykings model en die toepassing daarvan in
hipertensie studies uitgevoer in die Mamre gemeenskap en 'n hipertensie kliniek
populasie in Mitchells Plain.
| Identifer | oai:union.ndltd.org:netd.ac.za/oai:union.ndltd.org:sun/oai:scholar.sun.ac.za:10019.1/53739 |
| Date | January 2004 |
| Creators | Kepe, Lulama Patrick |
| Contributors | Lombard, Carl, Muller, Chris, Stellenbosch University. Faculty of Science. Dept. of Statistical and Actuarial Science. |
| Publisher | Stellenbosch : Stellenbosch University |
| Source Sets | South African National ETD Portal |
| Language | en_ZA |
| Detected Language | Unknown |
| Type | Thesis |
| Format | 148 p. |
| Rights | Stellenbosch University |
Page generated in 0.0027 seconds