August, 2014 - PhD Thesis Bangalore

Sampling Vs Non sampling Error

There are two types of error that we may find occurring when the effort is to try and estimate the parameters of the population from the sample. These errors can be classified as sampling and non-sampling errors. Sampling error: This kind of error is often seen arising when the sample of the study does not … Continue reading “Sampling Vs Non sampling Error”

Sampling error: This kind of error is often seen arising when the sample of the study does not represent the population that has to be studied. To understand better with an example, if the entire population comprises 200 MBA students of a business school and the research focus is to estimate the average height of these 200 students. The sample chosen is, let’s say, 10 students. In this case if we assume that the true mean of the population is known and the analysis show us that there is a wide difference between the sample mean and population mean. This kind of an error falls in the category of a sampling error. The reason for this kind of an error is the chosen sample size. In the above case, a sample of 10 is not a representative of the entire population. If the sample size is increased to 15 the error reduces. A significant increase in the sample size on one side significantly reduces the error on the other side.

Non Sampling error: This error arises because of various reasons. Some of the reasons are:

a) False or incorrect information given by the respondents may lead to a non-sampling error. For example, sometimes the respondent may not disclose his correct age and this may bring up a non-sampling kind of error.

b) Sometimes error arises when the transfer of data is being done onto a spreadsheet, from a manual sheet which is the questionnaire.

c) There are some errors that may happen at the time of coding or tabulation.

d) At times, it so happens, that the population of the study is not defined in the correct manner. It leads to errors.

e) The respondent that the researcher chooses for study, at times refuses to become a part of the study. This also becomes a kind of non-sampling error.

f) Another type of non-sampling error is the error of the sampling frame. Sometimes, the researcher decides to ignore a certain category of respondents and that may lead to the development of a non-sampling error.

Classifying Experimental Designs

Experimental designs should be categorized with many a variations. They can be classified and organised by understanding the application of the fundamental signal to noise ratio metaphor. This metaphor elucidates that what we see or observe can be split into two basic components. These two components are the signal and the noise. In most of … Continue reading “Classifying Experimental Designs”

In most of the researches, the signal has its link with the key variable of interest. The noise here comprises the random factors in the situation which make the visibility of the signal in the room relatively poorer. A ratio construct can be created when the signal is divided by the noise. When one talks of research, the signal should have high relativity to noise. For instance, if the treatment or programme and the measurement is also very good they can be termed as strong signal and low noise. In light of this concept, the experimental designs can also be classified into two categories. They can be termed as signal enhancers or noise reducers. Both these categories work towards enhancing the quality of the research. The first kind which is the signal enhancing experimental design is technically called the factorial designs. In this type of design, the entire focus is on the set up of the programme. It would help to examine and understand the different variations of a treatment.

In the other category, there are two major types of noise reducing experimental designs. They are called the covariance designs and blocking designs. The basic purpose of this kind of a design is to put the sample information and pre programme variables so that some noise from the study is taken out and more precise and worthy analysis can be done

Between Balanced and Unbalanced Scales

It is interesting and important at the same time to understand whether it is best to administer a balanced scale or an unbalanced scale. To begin with, if we focus on the balance scales, it primarily means that there an even number of points on either side of the scale. This means on the positive … Continue reading “Between Balanced and Unbalanced Scales”

On the contrary, a negative scale is something that is skewed on either the positive or the negative side. This is again an option to be used when the research scale had to be selected.

The balanced scale is bigger and gives more narrower options to the respondents, to chose between balanced and unbalanced scales, it largely depends upon the the kind of expectation or estimation that you have from the responses. If the researcher is aware that most of the respondents are going to agree to him the breakdown of the negative side of the scake will not be of much use and it is better to use an unbalanced scale. To explain it simply, in the case of the predicted skewed response on either side of the scale from the respondents, an unbalanced scale works well. However, if the researcher has doubt about the responses and cannot be sure if the responses will skew on either side then he should go for a larger scale like a balanced scale so that fair anount of clear options are offered to the respondents.

One point that should be noted in the case of unbalanced scale and should be considered thoroughly before deciding the scale is that, unbalanced scale do not offer a middle point and the result is an ordinal scale aas opposed to an interval scale. This makes it difficult to compute the mean or average response. And for the measures of central tendency,n the dependance comes on median instead of mean.

Considering all these factors, it becomes simpler for a researcher to chose between a balanced and unbalanced scale.

Cluster Analysis: A Classification Technique

Cluster Analysis is a grouping technique. This technique works on an assumption that states that the similarity is dependent upon multiple variables. It helps to measure the proximity of the study variables. The groups that emerge out of cluster analysis are homogeneous in their own composition and heterogeneous when it comes to comparison to other groups. The grouping for cluster analysis can be done for anything ranging from objects, individuals to products and entities. The researcher identifies a set of clustering variables. These variables are the identified variables that have a significant role in classifying the objects into various groups. For this reason cluster analysis is also called a classification or grouping technique. It has a lot of use in different branches of social sciences particularly psychology, sociology, management and engineering.

Cluster analysis is different from other data reduction techniques. The similarity of course is that it analysis the function of multiple independent variables but the difference is that, in factor analysis the original correlated variables are reduced to a more manageable number gut the data reduction is carried out on the columns of the data matrix. While, in the case of cluster analysis, the focus is on the rows which could be the individuals, entities, products or any other variable.

Another data reduction technique that can be confused with cluster analysis is the discriminant analysis. In the discriminant analysis the classification and identification of similarities is a pre requisite. It is imperative here to put across the objectives and rules of similarities in order. In the case of cluster analysis, the hole population is undifferentiated and all efforts to find out the similarity in the response to variables and the grouping task is done as an outcome of the cluster analysis.

The usage of cluster analysis is widespread and it has application in all the varied branches. It is the best classification technique when the factors involved in data collection are multiple. Its main use is seen in the segmentation technique where the main task is to split the potential customers within a market into different groups. Maximum explanation from the output of the cluster analysis has been witnessed in this field of segmentation.

Participatory Research: Methods and Practices

The participatory methods are focussed towards the planning and conducting the research process in the presence and involvement of people, those people whose life, world, actions and thoughts are under the process of research. As a consequence of this, the means and methods of enquiry and the different questions of research are created from the synchronisation of two different views and practices, involving both practice as well as science. Research bestows its benefits on both the sides. Practices that have been established as daily basis activities bring forth their own view point and perspective, always a new way to deal with a challenge. The participatory research process widens the horizons of the researcher and offers the freedom to co researchers to move beyond the limitations of standard routines, interactions in a more cognitive fashion. It helps to challenge and rework on established practices and interpretations. However, just by conducting participatory research, it is not ensured that there would be a convergence of science and practice. It is much more than that and a demanding process that is developed on two different spheres of action, both science and practice come into coherence with each other, interacts, develop lead to an understanding of each other.

The participatory method is an orientation of enquiry. Participatory research can be called a method that challenges concrete and fundamental research methods and supports the benefits of research that may lead to the involvement of the research partners to enhance productivity of knowledge. The approaches suggested for participatory research are not different and distinct from the standard empirical social procedures. The main characteristic of participatory research is it’s individuality and self-determination. It is not suggested and possible to canonize them in the form of a single, cohesive, methodological approach. Incorporation of processes is all the more important in participative research than in any other. If the purpose and desire is to gain a further deep understanding of the contextual structured of meaning and the dynamism that in embedded in social action, it becomes worthy to incorporate participatory research at the design stage itself.

To conclude, participatory research does pose certain questions in context to knowledge and research in a very radical fashion. It is so strong an area of research that it has the capacity to bring attention to the neglected areas of the methodology and kindle their further development.