Geri Dön

Pazarlama araştırmalarında optimal ölçekleme yöntemleri ve uygulanaları

Optimal scaling methods in marketing research and application

  1. Tez No: 66756
  2. Yazar: BAHAR ÇELİK
  3. Danışmanlar: DOÇ. DR. BURÇ ÜLENGİN
  4. Tez Türü: Yüksek Lisans
  5. Konular: Mühendislik Bilimleri, İşletme, Engineering Sciences, Business Administration
  6. Anahtar Kelimeler: Belirtilmemiş.
  7. Yıl: 1997
  8. Dil: Türkçe
  9. Üniversite: İstanbul Teknik Üniversitesi
  10. Enstitü: Fen Bilimleri Enstitüsü
  11. Ana Bilim Dalı: İşletme Mühendisliği Ana Bilim Dalı
  12. Bilim Dalı: İşletme Mühendisliği Bilim Dalı
  13. Sayfa Sayısı: 148

Özet

ÖZET Bilgi çağında olsak da; insanlar sürekli karşılaştıkları bilgilerin işlenmesi, tasnif edilmesi ve yorumlanmasında son derece zorlanmaktadırlar. Pazarlama araştırmasında da benzer şekilde toplanan verilerin tasnif edilmesi, analizi, yorumlanması gittikçe daha fazla önem kazanmaktadır. % analizleri, basit regresyon analizleri çeşitli şekillerde elde edilen çok fazla verinin yorumlanmasında kimi zaman yetersiz kalmaktadır. Ağırlıklı olarak çok yüklü matematiksel hesaplamaların bulunduğu araştırma tekniklerinin; - örneğin bu tezin kapsamına giren correspondence, homogeneity ve doğrusal olmayan temel bileşenler analizi gibi -, ise teorisi 1930'lu yıllarda geliştirilmişken uygulamanın karmaşıklığı ve zorluğu sebebiyle bu tür çalışmalara yaklaşık 1970'lere kadar ara verilmek zorunda kalmıştır. Ülkemizde de faktör analizi, kümeleme ve çok boyutlu ölçekleme gibi analizler pazarlama araştırmasında 80'li yıllardan itibaren yoğun olarak kullanılmaya başlanmıştır. SPSS gibi yazılımların da bu analizleri kapsamı dahiline alması, tüm dünyada ve Türkiye'de bu sürecin hızlanmasında rol oynamıştır. Günümüzde ise nominal verilerin daha detaylı analiz edilmesi, analiz sonuçlarının görsel olarak sunulabilir olması önem kazanmıştır. Adlı ölçekle ölçülmüş verinin sayısal yöntemlerle analiz edilmesini, değişkenin kategorileri arasındaki doğrusal olmayan ilişkilerin belirlenmesini ve değişkenler arasındaki benzerlik ve farklılıkların hem sayılarla hem de görsel tekniklerle ortaya konabilmesini sağlayan bu analiz tekniklerinin uygulamada çok yarar sağlayabileceği düşünülmektedir. Bu çalışma kapsamında optimal ölçekleme yöntemlerinden üç tanesi olan correspondence analizi, homogeneity analizi ve doğrusal olmayan temel bileşenler (non-lineer principal components) analizi incelenmektedir. Bu analizler kendi içlerinde farklı gibi görünseler de aslında kullanıcı için birbirinin devamı veya ikame analizler olarak kullanılabilmektedir. Bu nedenle çalışmanın uygulama kısmında da aynı veri setinden yararlanılmıştır. Bunun amacı analizlerin birbiriyle karşılaştırılmasını kolaylaştırmaktır. Optimal ölçekleme yöntemlerinden dördüncüsü olan doğrusal olmayan kanonik korelasyon (non-lineer canonical correlation) analizi ise diğer üç analizden farklı olarak tek veri setini değil birden fazla veri grubunun karşılıklı etkileşimini analiz etmektedir. Doğrusal olmayan kanonik korelasyon analizi yapı itibarıyla diğerlerinden çok farklı olduğu için bu tezin kapsamı içine alınmamıştır, istenirse bu analiz ayrı bir tez konusu olarak incelenebilir. X-

Özet (Çeviri)

SUMMARY OPTIMAL SCALING METHODS IN MARKETING RESEARCH AND APPLICATIONS Although we are in a century of information, people have difficulties in handling, classifying and interpreting data. Likewise in the field of marketing research the classification, analysis and interpretation of collected data become more and more important, x2 analysis or simple regression analysis are sometimes not sufficient to interpret a bunch of data which is collected in various ways. Research techniques which predominantly include heavy detailed calculations (e.g. correspondence analysis, homogeneity analysis and non linear principal components analysis in the framework of this thesis) are developed in 30's but could be continue only in 70's because of the complexity and difficulty of application. People must use computers and software programs in certain analysis because for example eigenvalue and eigenvector techniques make use of intensive matrix transactions and principal components analysis or alternating least squares technique makes use of iterative methods. Also in our country, analysis such as factor analysis, cluster and multidimensional scaling have been used intensively in marketing research ever since 80's. The fact that software like SPSS include such analysis contributed to the development of this process in Turkey and all over the world. Nowadays, more detailed analysis of nominal data and visual presentation of the results have become more important. While deciding for these thesis, first we controlled the software programs. We found out that SPSS 6.1 Categories for Windows contains all four analysis techniques under the umbrella of optimal scaling..XI-These techniques of analysis provide the quantitative analysis of nominal data, the determination of non-linear relations between variable categories, quantitative and visual expressions of differences and similarities between variables. Therefore, they will probably be very practical. In the framework of this study three optimal scaling techniques, i.e., correspondence analysis, homogeneity analysis and non-linear principal components analysis are examined. Even though they seem to be different from each other, actually they can be utilized substitute or complementary analysis by a researcher. That is why, the same set of data is used in application section of the study. The aim is to make the comparison of the analysis easier. In the application section by adding new variables to the data sets when necessary or by changing the measurement levels of the variables each technique of analysis is exemplified in the most suitable environment. In that way, the best conditions for the comparison of similarities and differences between these three techniques of analysis have been created. All of these three analysis belong to the group described as dimension reduction in the framework of multivariate analysis. That means the relations between variables are represented by only a few dimensions (e.g. two or three). For instance factor analysis which is based on principal components analysis is included in this group. This enables you to describe structures or patterns in the relationships between variables that would be too difficult to fathom in their original richness and complexity. In marketing research applications, these techniques can be a form of perceptual mapping. A major advantage of these procedures is these procedures is that they accommodate data at different levels of measurement. In standart statistical analysis, level of measurement is a fixed property of each variable in the analysis. The measurement level guides the choice of an appropriate technique. In optimal scaling, level of measurement is usually regarded as a user-specified option. By adjusting the specified level of measurement of some variables in the analysis, you may be able to uncover hidden relationships. -Xll-For our purposes, there are three levels in optimal scaling: 1 The nominal level 2.The ordinal level 3.The numerical (interval) level The fact that a variable is intrinsically numerical does not mean that a relationship with another numerical variable has to be linear. Two numerical variables can have a nonlinear relationship. For example, age in years and house spent at work can both be measured at the numerical level, but since both children and retirees spend little or no time at work, the linear correlation between these two variables will probably be low. Optimal scaling can detect nonlinear relationships and produce maximum correlations between the variables in your analysis and in your research. The four optimal scaling procedures - correspondence analysis, homogeneity analysis, nonlinear principal components analysis and nonlinear canonical correlation analysis- extend the classical statistical techniques of principal components and canonical correlation analysis to accommodate variables of mixed measurement level. If the variables in the analysis are all numerical and relationships between the variables are assumed to be linear, then standart correlation- based statistical procedures in SPSS should be used and there is no need to turn to optimal scaling procedures. However, if variables in the analysis have mixed measurement levels, or if nonlinearities in the relationships between some pairs of variables are suspected, then the appropriate optimal scaling procedure should be used. As mentioned above, research techniques which predominantly include heavy detailed calculations for correspondence analysis, homogeneity analysis and nonlinear principal components analysis are developed in 30's but could be continue only in 70's because of the complexity and difficulty of application (Van Der Burg, De leeuw, Verdegaal, 1988, S. 178). The first one is the correspondence analysis which quantitatively and visually demonstrates the similarities and difficulties between the categories of two variables that are analyzed at nominal level. In practice, it is frequently used in brand preference or positioning studies. - Xlll ?This technique can be used in any case where cross tabs in marketing research are employed. Especially in variables such as occupation, car brand or banks which require too many categories, it is very to employ and interpret cross tabs. Yet correspondence analysis provide visual presentation of these relations in its joint graphic and in that way eliminates the disadvantages of cross tabs. In this analysis, the researcher has four options for normalization of data. The generally adapted technique is canonical normalization. It is the most suitable one in terms of demonstration of the similarities and differences between both row and column scores. In the context of this study, the application correspondence analysis to any cross tab which is not based on any data matrix. In some cases, the researcher may need an additional third variable, Just like the cross tabs, the correspondence analysis cannot offer solution to this problem. However, homogeneity analysis which is another optimal scaling technique, provides a solution. Homogeneity analysis detects the similarities and differences between variable categories, as the correspondence analysis can do, for three or more variables. Unlike the correspondence analysis, homogeneity analysis does not demonstrate row and column scores graphics separately. However, this analysis includes the graphic discrimination measures, indicating the size and degree of the representation of each variable, and the graphic object scores indicating the distribution of all observations within the predetermined axis. The graphic showing the distribution of the categories of all the variable can be interpreted in the same way as the correspondence analysis. The object scores graphics shows the number of observations which are used as a base for interpretations. To check the reliability of an interpretation, the researcher can refer to the object scores graphics which shows the number of observations related to that interpretation. In all of the three analysis, this matching can be easily performed since the predetermined axis are the same for all the graphics. -XIV-If the homogeneity analysis exercises to explain the relations between only two variables, the interpretation will be similar to that of the correspondence analysis but results are not identical. These two analysis have the same purpose but they are based on different calculation methods. While correspondence analysis is close to %2 analysis, homogeneity employs alternating least squares method. Nonlinear components analysis applies on three or more variables just like homogeneity analysis, however the measurement level of variables can be changed, This is the main difference of this analysis from the other two. While in correspondence and homogeneity analysis the variables are examined only at nominal level, in non-linear components analysis the measurement level can be raised by choosing one of the ordinal, interval or numerical measurement levels. This option helps to detect the increasing or decreasing tendency of a variable. In that way different kinds of hypothesis can be tested which make interpretation beginning with“as the age increases”or“as the education level decreases”possible or provide the analysis for ordinal, interval and numerical variables with more than on nominal variable. Non-linear components analysis offers graphics similar to homogeneity graphics. Therefore, the interpretation of the results can be performed in similar ways. Alternating least squares is the base of these two analysis. Despite this fact, in terms of measurement limitations, correspondence and homogeneity analysis seem closer to each other. That is the reason why in the literature the homogeneity analysis is sometimes referred to as multiple correspondence analysis. If all the variables in the non-linear components analysis are nominal, the results will be identical to those of the homogeneity analysis. Likewise, if all the variables in the analysis are numerical, the results will be equivalent to those of the factor analysis which utilizes the option of standard principal components analysis. All of these three analysis can be used just like in the application section. At the same time, the researcher can make use of them in analysis like segmentation, life styles, purchase tendencies, brand preference or product/ brand positioning. In marketing research or in professional life, in addition to the above mentioned researches, these techniques can be employed as a form of perceptual mapping. -XV-Unlike these three optimal scaling techniques, the last one, which is called non-linear canonical correlation analysis does not analyses one set of data. It deals with the relation between two or more data sets. Non-linear canonical correlation analysis is structurally different than the other three optimal scaling techniques. That is why it is not included in the framework of this thesis. It can be dealt as a topic of another thesis..XVI-

Benzer Tezler

  1. Tez No

    909071

    Hibrit döşeme kirişleri: Tarihsel gelişim, tasarım örnekleri ve performans değerlendirmeleri

    Hybrid floor beams: Historical development, design examples and performance evaluations

    TUĞÇE SOYTAŞ İNCİLİ

    Yüksek Lisans

    Türkçe

    Türkçe

    2024

    Mimarlıkİstanbul Teknik Üniversitesi

    Mimarlık Ana Bilim Dalı

    PROF. DR. OĞUZ CEM ÇELİK

  2. Tez No

    26028

    Kar amacı gütmeyen kuruluşlarda pazarlama faaliyetleri ve bir belediye örneği

    Başlık çevirisi yok

    HASAN AYYILDIZ

    Yüksek Lisans

    Türkçe

    Türkçe

    1993

    İşletmeİstanbul Üniversitesi

    Pazarlama Ana Bilim Dalı

    PROF. DR. MUHİTTİN KARABULUT

  3. Tez No

    335595

    Broker-based ad allocation in social networks

    Sosyal ağlarda acenta tabanlı reklam atama

    İZZEDDİN GÜR

    Yüksek Lisans

    İngilizce

    İngilizce

    2013

    Bilgisayar Mühendisliği Bilimleri-Bilgisayar ve Kontrolİhsan Doğramacı Bilkent Üniversitesi

    Bilgisayar Mühendisliği Bölümü

    DOÇ. DR. HAKAN FERHATOSMANOĞLU

    YRD. DOÇ. DR. BUĞRA GEDİK

  4. Tez No

    43071

    Mamül hayat döneminde izlenen pazarlama stratejileri ve EMSAN madeni eşya emaye sanayi ve ticaret A.Ş'de bir uygulama

    Başlık çevirisi yok

    ZEYNEP ÖZKOCAER

    Yüksek Lisans

    Türkçe

    Türkçe

    1995

    İşletmeAnadolu Üniversitesi

    PROF.DR. BİROL TENEKECİOGLU

  5. Tez No

    900044

    Siyasal pazarlama perspektifinde seçmen tercihi ekollerinin entelektüel düşünme etkisinde karşılaştırmalı analizi

    Comparative analysis of voter preference schools on the effect of intellectual thinking in political marketing perspective

    ESAT SAYGIN

    Doktora

    Türkçe

    Türkçe

    2024

    Halkla İlişkilerAnkara Hacı Bayram Veli Üniversitesi

    İşletme Ana Bilim Dalı

    PROF. DR. METEHAN TOLON

Geri Dön