Hello and welcome to this video on Primary Quantitative Data Analysis. In the previous week's video on Data Collection Methods, we learnt various ways that you can collect data. This included both primary and secondary data collection methods. Primary data collection basically entails conducting a survey to collect first hand information based on your research. In this video, we will look into the various steps of primary data preparation and inspection and after the video, you will be able to understand how to prepare your data for statistical analysis. Before we proceed, you need to be sure that the amount of questionnaires administered are enough to perform statistical analysis. Do you remember the calculation of a sample size? If not, take a look at the video about sampling and the week four. Secondly, the way in which you prepare the questionnaire is of utmost importance as a tool determine how you carry out the analysis based on the measurement levels of your indicators. You may refer to the video on formulating and administering a questionnaire also under week four to refresh your memory on constructing a questionnaire. Now, with reference to data preparation and analysis, I like to use the analogy of getting a vehicle ready to go on a race. There are a number of things that you should ensure are working in perfect condition such as making sure that the tank is full, checking that the steering wheel is in perfect condition, making sure that the seat belt is functioning, and so on. In this case, the car is the data preparation and the race is the data analysis. This will become more clear as we go on, but for now, picture the racing car and the inspection that you need to perform. I will try to simplify data preparation by highlighting the important steps to take in primary data preparation. Let's assume that you have already collected your primary data. The first step will be to convert the primary data into a format that can be used in the statistical software that you intend to use for analysis. This may be any software such as SPSS, STATA, R, and so on. You may well know that in primary research studies involving the use of participatory tools, much of the information gathered is of qualitative nature. Quantitative methods of data analysis, however are of great value to the researcher who is attempting to draw meaningful results from a large body of qualitative data. As a researcher, you can have quantitative analysis approaches that can be applied to qualitative data. For example, suppose the interest of the researcher is to learn about people's perception on quality of life. It is likely that the explanation that results from the discussions across several cultures will show some frequently occurring answers such as work-life balance, strong friendships and relationships, high level of income, and so on. Such information can be taken from the explanation and quantified for analysis. This involves coding the qualitative data into different numeric values. Quantitative approaches provide the opportunity to study this coded information first and then turn to the remaining qualitative components in the data. Once you have coded the data, you can analyze it using the statistical software. For the purpose of this session, we will make reference to the use of SPSS software, but bear in mind that there are other statistical packages for analysis of data as I have briefly mentioned before. With reference to SPSS, there are different ways to inputs the primary data. One way is to inputs the data directly into the SPSS database, and another way is to inputs the data into an Excel workbook, and then import the data based for use in SPSS. There are many online tutorials available to learn how to import data in SPSS. The next step would be data inspection to check whether the data input is done correctly. Do you remember the inspection of the car before going on a race? The car here is the data that we need to inspect before embarking on the analysis. Bear in mind that it is quite common to have a situation where all your questions are not answered. And in this case, there may be a number of missing values in the dataset. The missing values cannot be analyzed, but you need to assign them a value in your database. This is particularly important as it would interfere with the mean score of the data points in your variable. Usually, you assign a code to the missing values. The code assigned to the values should be one that has no possibility of occurring anywhere in the variable points. Let me elaborate this further. For example, the most commonly used value is 999 or triple nine. This would be fitting for data that is measured on a likert scale or data with low value ranges where the minimum value is say, one, and the maximum value is five. 999 would then in this case, stand out as a data point that is unusual. This however, should be taken with caution if you have variables that may contain continuous values greater than 999 such as income data. And perhaps, another value such as minus 999 or negative 999 can be assigned. Another important step in data preparation is computing variables. With reference to the conceptual framework, you develop concepts from theory, which are measured by variables. Variables can be further measured using indicators. For example, taking social status as a concept may be measured using a dimension such as wealth. Wealth could be measured by the assets, accumulated, or perhaps, the amount of disposable income. When collecting the information by survey, the questions are indicators that you eventually compute into a variable. Explaining your results at variable level makes it more understandable and helps you to report and answer your research question to the point. Now, before completing the indicators into variables, you need to perform a reliability analysis to check for the level of consistency that allows for the indicators to be computed into one variable. It is important here to note that the measurement of the indicators should be the same. What I mean is if you are using a five point likert scale, for example, all the indicators that you intend to compute into one variable should adopt the same likert scale. Again, for reference to the measurement of the indicators, refer to the video on formulating a questionnaire under week four. As I mentioned, the reliability analysis is performed before computing the indicators into one variable. The output of the reliability test is a Cronbach's Alpha, which is an indicator or a coefficient of consistency. It estimates the reliability of a variable. The coefficient ranges from zero to one where zero indicates no consistency in the measurement of the variable and one indicates perfect consistency. At least, three indicators are required to calculate the Cronbach's Alpha. Now, the question would be how do I determine the consistency of the variable if the values range from zero to one? Well, the rule of thumb is 0.7. Anything below that requires that the researcher checks for improvements and perhaps, by reducing the number of indicators that are grouped together, but still bearing in mind that you require three or more indicators to compute the variable. After making improvements, values between 0.6 and 0.69 is doubtful, but acceptable for thesis research. Values below 0.6 means that your variable is not measured in a consistent way and is therefore, unreliable. In this case, you will have to report the results using the indicators. In conclusion, you now have a basic idea of some of the important steps in data inspection, but we cannot yet run our car for the race. In the next video, we will look into more profound data inspection methods that are necessary before performing inferential statistics.
