---

---

---

---

---

---

Chapter wise analysis of Shodhganga SGRRU Dehradun PhD Thesis critique

Chapter 4: Materials & Methods

What happens is that when we write a PhD thesis, most researchers tend to become authors and forget that their primary and key objective is to describe their research and not write a book. Yes, the research description should be such that a layman could understand it, but the term \”layman\” when used in context of science communication through written documentation, is a person who should know the basics of research.

Let me elaborate, In this chapter the scholar, in his first paragraph, has tried to describe research methodology and research design in general; but has failed to talk about these in context of this particular research work. In contrast, the second paragraph talks specifically about target population and sampling specifically in the context of this research. This type of writing begs us to conclude that the researcher understands the concept of sampling better that the concept of research methodology and research design. Which, in general is the case of most of the PhD scholars in India.

Although, research design for this specific research has been described, the description itself is organized, haphazardly to accommodate all the points that need to be communicated. The writing is not logically sequenced, with some points overlapping and out of sequence. Vocabulary, is off, example: improper use of the word \”validated\”. Proper reference either internal or external with respect to the document is missing, example: \”As the scales have been already identified\”.

Within the document, on many instances, some words are missing, insinuating automated text formatting techniques, lack of proofreading and evaluation, example: \”the influence of the it on its overall\”.

Hypotheses five to eight, although implicit from the topic of research were absent from the stated objectives and research questions. Thus seem to have been an after thought and might have emerged or had revealed themselves to the researcher, when they were drawing figure 4.2 \”Proposed conceptual model of study\”.

Six items associated with country of origin do not explicitly mention the countries in \”Countries of Origin\” but proxy them with a characteristic that one might associate with that specific country. The reliability of the scales is to be taken at its face value without knowing the names and credibility of the \”experts\”.

Choosing the final sample size is arbitrary. Researcher references other studies having a near to absolute quantum of sample. Convenience sampling is chosen because some other researcher said that convenience sampling is the most \”dominant choice available when it comes into conducting quantitative research\”, this statement is absolutely absurd in itself. Although a justification for determining a sample size is presented, it is not scientific or statistical in nature. Unfortunately, again the said pilot with 58% response rate from 200 sample, has not been used to determine the sample size and statistical power. This forces me to conclude that that either a pilot was not conducted or guides / supervisors themselves do not know how to use the findings from the pilot to determine the appropriate sample size of the study.

Since, responses were collected anonymously and through google forms, it will be hard to conclude that the actual responses were from people with the correct demographics, rendering this research unusable / non-reliable.

The statement \”The issues of data accuracy comes when the data is collected in an off-line mode.\” as made by the researcher shows the level of ignorance while collecting data thorough on-line modes and through self-administered Likert scales.

While univariate distributions can provide useful insights and preliminary checks for normality, they cannot fully determine multivariate normality. A comprehensive approach should include multivariate tests and analyses to assess the joint distribution of the variables. Therefore, the statement \”Consequently, the investigation of multivariate normality was conducted through the use of univariate distributions in this research study.\” shows the halfheartedness, the researcher has towards this research and its rigor.

In defining and explaining univariate normality, the researcher has sited existing literature, however has failed to explicitly list out as to what and how he will be using the same in his research. which, in-turn begs us to conclude that this paragraph is generic and is not updated or editited to accommodate this specific research. Moreover, this is seen to be in contrast of the next paragraph on Outliers.

The researcher says \”Analysing the entire range of bivariate scatterplots isn\’t feasible\” and cites a 1983 paper for the technique involved, today numerous software tools are available for automating this task which were not available in 1983 (more than 40 years). This, again shows the rigour of our research.

Again, while writing on test of multi-collinearity and singularity, the researcher has failed to communicate what he has used of this specific research and describes the concept in general.

The researcher says \”Statistical considerations were the basis for selecting Cronbach\’s α\”, but fails to elaborate on them.

Chapter 5: Results and Discussion

Results of examining for missing values, outliers, normality, linearity and multicollinearity have not been presented, only a citation is given.

More than 75% of the data that was analyzed came from the age group of 18-25 years, which is far lees (at-least 10 years) than the average car buying age in India (mid 30s) let alone the location of the study. This, in-turn renders the inputs of (93% less than 35 years) respondents as largely irrelevant to the key objectives of the study.

Bar charts have not been arranged properly to communicate the results. Marital status, education and annual household income as chosen variables is irrelevant to the study objectives. Since, 70% of the respondents are students and below the age of 25 years.

There is a disconnect between table numbering on tables and their reference in the text.

The construct \”Country of Origin\” is not normally distributed since both the skewness and kurtosis are far away from zero across all six items. This suggests two things, either the questions were not framed around having a zero value and / or the scale was not centered around zero value. But, there is consistency across all the six items. A graphical representation is missing. Similar is the case with \”Brand Association\”.

Inconsistency is observed across the five items of the \”Brand Loyalty\” construct. For the construct \”Perceived Quality\”, PQ4 is the most normally distributed item. In case of \”Brand Awareness\” BAWA6 is the most normally distributed item.

Shapiro-Wilk test results have been displayed for the constructs and not the items within the constructs. This raises considerable questions on normality of the data since previous analysis largely opposes normality. Furthermore, if individual item values were consolidated to construct level values, the same is not presented.

When using multiple items to measure one construct through a Likert scale, it is imperative that these items HAVE collinearity, because all the items are intended to measure the same construct. In this study, however, the researcher finds \”no indication of multicollinearity in the data set\”. This, indeed indicates that the questions / items were not framed correctly or even worse, the data set is fabricated / manipulated. This same point is in opposition to the findings presented under Convergent Validity sub-section.

In order to test Inter Construct Correlations, the researcher has not given appropriate methodology deployed to transform item level values to construct level values.

Significance level of 0.000** in Multiple Regression Analysis sub-section is highly questionable. Similarly, p-values of 0.0000 in Direct Testing Effects sub-section are also highly questionable.

The three paragraphs across pages numbers 178-179 are in opposition to the findings of this study, as if study was done by someone else and was written by someone else.

Study\’s findings of brand loyalty as a mediator for country of origin effects on brand equity are incorrect as there is no evidence of direct effect and p-value is 0.000 for indirect effect. Secondly, 42% of the respondents prefer Indian origin brands and the respondents are also of Indian origin. Which, in-turn will bias brand loyalty with national loyalty. The findings will thus be different if these 42% of the responses were removed from the analysis. Same would be the case for Brand Association and Brand Awareness.

In-order for this study to be relevant, one has to obtain real world brand equity values for said brands and compare the predictions / predictors / prediction models. Lastly, this study shows how confounding variables (country of study and national loyalty in this case) can affect research findings significantly.

---

---

Data Analysis Health Score (DAHS) for Shodhganga SGRRU Dehradun PhD Thesis critique,

The overall data analysis health score for this PhD thesis evaluates to be -52%, as shown in the graph above. Furthermore relative scores for individual analytical processes are also indicated in the above graph, and are as follows. The thesis scores 80% on Framing and Objectives, -83% on Data Integrity & Provenance, -46% on Exploratory and Descriptive Statistics, -85% on Inferential Statistics, -100% on Predictive and Prescriptive Modeling, -33% on Reproducibility & Transparency, 0% on Ethical and Responsible Conduct and -100% on Reporting & Documentation.

---

---