Measurement And Instruments For A Quantitative Research Plan
Levels of measurement used in the research
When carrying out research in social sciences, one needs to be reliable and accurate. Data collection in social sciences takes so many forms. These includes measurement of cognition, perception, opinions and others that cannot be measured directly. In the quantification of perceptions, events and people, there are four types of measurements that are majorly used. (Miller & Salkind, 2002)
Read also Levels of Measurement and Concept of Validity
The following are the four main levels of measurements that I found to be relevant for my study. I found these methods important because they match the type of data I collected and how I will use them in the analysis and finding the results. These thus goes hand in hand.
The four levels of measurements.
The nominal scale
This is also referred to as dummy codding. This method works by placing people, perceptions, events and many others into categories basing on some common traits. Some of the data naturally suits in the nominal scale. Examples of such include Americans vs. Asians, male vs. females, redheads vs. blondes and many more. The nominal scale is the basis in which analyses such as Analysis of Variance (ANOVA) are formed as they require comparison between two categories. The nominal scale in this case falls into the lowest form measurement as it does not capture any information related to a focal object but basically groups the objects into categories. Coding in this case is done by use of numbers, labels or any symbol that can best represent the category that an object or a person belongs.(Miller & Salkind, 2002)
This type of scale has got one major advantage over the nominal scale. It has all the details that are captured in the nominal scale but then goes ahead to rank the data collected from the lowest to the highest. They give an idea of where the data lies in relation to one another. The ordinal scale is evidently richer than the nominal scale but suffers information loss as it only ranks without giving more info on how far apart the ones ranked are. (Trochim & Donnelley, 2001)
Unlike the two discussed scales of measurement, interval scale provides richer information about an object being studied. It denotes the distance one object is from the other thus providing more information about it. (Isaac & Michael, 1971)
This is the scale that provides the richest information about the object. This type of scales has all the information that all the previous three scales have but also contains an additional absolute zero point.(Trochim & Donnelley, 2001)
These four levels of data collection discussed above have an effect on how data is collected and analyzed later. Data collected wrongly will cause an adjustment to the analyses, design and basically the whole research. As I mentioned before, this is the main reason why I chose the above mentioned levels of measurements as they matched my levels of data collection.
Content, empirical and construct validity
Validity is the determinant to whether or not a design is well designed or not well designed and gives the outcomes that seen to be suitable to generalize the population of interest. (Cozby, 2001)
Construct validity refers to the degree to which a particular test measures what it is required to measure. It is very essential to the recommended validity of a test. (Bagozzi, Yi & Philips, 1991)
Read also Concepts And Types Of Reliability And Validity That Apply To Tests
Empirical validity also known as predictive or statistical validity illustrates how close the scores in a particular test correlate with the behavior as studied and measured in other contexts.(Cozby, 2001)
Content validity also referred to as logical validity describes the extent to which a particular measure stands in place of all the other facets in a particular social construct. (Cronbach, 1971)
How to ensure the three types of validity in a study
It is normally assumed that the study is valid just because the study carried out is scientific. This is normally not true. The researchers who carry out the scientific study are normally pushed by external forces such as the desire to get some certain results. Due to unreliability of the scientist, it is important to ensure that the results are reliable and conclusive. A reliable study avoids biases, utilizes the recommended sample size and majorly use random sampling procedure to collect data.(Cronbach, 1971)
Read also Validity and Reliability – Compare and Contrast Essay
To ensure validity, the following has to be undertaken;
This is critical in ensuring validity of any research. It may be by the use of a random number generator or by use of a computer to collect data. This ensures that there is no bias. It does so by producing the comparable groups such as in the terms of age, gender, the participant characteristics and many more key factors.(Cronbach, 1971)
The desired population should be able to carry out the study to the conclusion of it.A sample population is thus taken to represent the population. It is thus very important to get a reliable sample size so as to achieve reliable and statically significant results. (Moskal & Leydens, 2000)
Bias in results
This involves the production of findings that should not be produces due to alteration of methods. The most common forms or types of biases include; intervention biases which occurs when there is a difference in how the subjects were prevailed to the matter of interest, measurement biases which may be caused by social desirability where people favors themselves and thus may fail to provide honest responses and finally selection biases which occurs when a certain sample is omitted purposively.(Cronbach, 1971)
To improve the validity the following has to be ensured:
- There should be clear definition of the goals and the objectives
- The assessment measure should be matched to the goals and the objectives
- Comparison of data should be made to ensure accuracy (Cozby, 2001)
How to ensure the reliability of the measurements in the study
Reliability is basically the level to which a particular assessment too is capable of providing consistent and stable results.(Cronbach, 1971)
The following will be used to ensure reliability in this text;
- The test-retest reliability. This is basically a type of reliability achieved when one administers the same type of test twice to the same group of individuals over a certain period of time.
- Parallel forms reliability. This is achieved by issuing the different versions of a particular assessment to the very same group of people. The different scores can thus be correlated so as to determine how consistent the results are. (Morse, Barrett, Mayan, Olson & Spiers, 2008)
- Inter-rater reliability. This is used to assess the level in which the different raters or judges agree in their assessment. This is beneficial in that different people or judges in this case will have different views thus making it very reliable as different interpretations are made. (Moskal & Leydens, 2000)
All the above mentioned types of reliability are capable of determining the reliability of a measurement in a study.
Strengths and limitations of measurement using questionnaires
- Very effective way of measuring people’s behavior, preferences, attitudes and opinions thus reliable
- They enable replication thus making it easy to check for reliability
- Questionnaires are distributed to several people in an area thus making it easy to gather opinions from different groups of people. (Munn & Drever, 1990)
- They can be given to the same people twice so checking for reliability is easier.
- The different versions of the questionnaires can be made and the correlations made to assess the consistency of the answers given. (Kimberlin & Winterstein, 2008)
- Respondents may end up lying due to social desirability as they want to portray a positive image
- People may answer wrongly due to language barrier or misunderstanding the language.
- The closed ended questionnaires are not detailed thus creating a lesser scope for the assessment which makes it unreliable.(Munn & Drever, 1990)
- The open ended questionnaires are not suitable for data collection and analysis as they need the researcher to read them in detail.
- They are not suitable for the less educated as they may end up giving wrong answers. This is because it requires superior writing skills and understanding of the questions so as to express the answers well.(Kimberlin & Winterstein, 2008)