It is important to understand statistics whether you are a student or a professional as more and more companies and businesses are incorporating statistical analysis in their strategies of management. This process can be overwhelming especially if you have no prior knowledge of conducting said analysis. Most students utilize statistical analysis models in their research papers. It is important to understand that data analysis is not only collecting the data but being able to spot and interpret trends, and explain your findings accurately. In this fast paper, we’ll look at the dos and don’ts when attempting to perform a statistical analysis particularly for a research paper.
Statistical analysis is a mathematical discipline in science that involves the collecting and organizing of data that is then analyzed, interpreted and presented in order to discover and explain trends or patterns. Statistical analysis is a vital research tool that is used mostly by researchers, scientists, governments, companies etc. Statistical analysis aids these organizations in decision and policy making.
1. Relevance
The statistical analysis you intend to execute has to be of relevance to your research subject. For instance, if your research topic is the impact of vaccination on covid-19 outbreaks in the UK, then the data you collect has to be on the population of the UK, the number of those vaccinated, number of positive covid-19 cases before and after vaccination etc. Basically the data has to revolve around your research topic. In this case you cannot collect data on how many people have cable TV as it would be redundant and irrelevant. Always ensure that you remain on course.
2. Go through your collected data
While in the collection stage, you maximize on variables and subjects. Before you embark on analyzing your collected data, it is important to get rid of unnecessary information, read data. Data that does not fall in the category of what is needed or appears to differ significantly from other variables or observations is called an outlier. An outlier may come from an error in collection or during the experiment or even from the difference exhibited by data points in a data set. In most cases the outlier is gotten rid of from the data set.
3. Understand your research type
There are different methods of conducting a statistical analysis. Each type of research will require a different technique to be employed for data analysis. There are 2 types of research; qualitative and quantitative research.
Qualitative research uses descriptive and thematic analysis. They employ data collection methods such as interviews, questionnaires etc. No software is needed for these types of research data. Quantitative research however involves numbers mainly and require software such as SPSS,R to analyze it. The researcher cannot make conclusions based on their understanding but instead have to rely on the results that they receive from the software. For accurate results always ensure you utilize the right analysis techniques.
4. Make use of assumptions
Every parametric test in statistical analysis assumes a particular characteristic about the data. These assumptions are what guide the researcher when executing the analysis. The assumptions vary depending on the parametric tests. The most common assumptions are normality, skewness, homogeneity, Shapiro Wilk’s test etc. The researcher should be able to know what assumptions to use in their research. The assumptions have to be followed to facilitate accurate interpretation of the data. If these assumptions are violated, then the interpretation and conclusion consequently change.
5. Don’t wing it
For you to avoid wasting time and collecting data that you will not need or use, ensure you have a plan beforehand. It is important to understand what questions you want to answer, the hypotheses you want to test for and which parametric tests you will have to conduct to get reliable results. Hubspot.net advises to have a written down plan detailing the whole research process. This will make it easier to carry out the analysis correctly.
6. Do not P-hack
P-hacking is a confirmative inflation bias, in simple terms, it is selective reporting. This occurs when a researcher intentionally leaves out some results after analysis because they do not give the preferred result in a bid to influence the outcome of the analysis, hence misreporting the true effect sizes in data analysis. They basically choose to report the tests that produce statistically significant results that agree or confirm their hypotheses.
Cases of p-hacking occur when you increase a sample size and then reanalyze to get a better result, remove outliers, try a different test or method of analysis in a bid to get different results etc. Always take the results as they are and steer clear from trying to make alterations.
7. Choose a sample size
Always ensure that you choose a good sample size. It should be just right, not too big and not too small. Whatever sample size you settle for ensure that you stick with it. Manipulating sample sizes to give you the results you want is tantamount to p-hacking and misrepresentation of the reality.
It is very normal to get weird results once in a while. Data analysis, especially big data in quantitative research, is no walk in the park. You have to be prepared to fail in analyzing statistics. Always review your techniques in the event that you get incorrect results. Feel free to contact this quick writing service for help in your analysis.