So far, I have also helped to do the data analysis of thousands of student papers, including the writing of some whole papers. Because I do market research and data analysis, my main tool is spss. I can't say that I am 100% proficient in spss, but it should be enough to cope with eighty or ninety. Naturally, I usually use my spare time to help students do some paper data analysis and paper writing guidance.
The core of many papers contains data analysis, and statistics should also be an important course in all disciplines. On the contrary, many disciplines only regard statistics and data analysis as an elective course, or even an unimportant course, which leads to students' complete incomprehension at the end of the paper.
In this case, many students know nothing about data analysis, which leads to problems from the initial design to the later data collection and collation, and finally leads to problems in analysis.
So, how to build a paper and write it from scratch when you don't know anything about data analysis? Although the data analysis part is the core of many papers, no matter what kind of paper you write, you can't do without the framework of the paper. Therefore, the specific process should be as follows:
The first is the topic selection, of course, many times the tutor directly gives the topic selection without much discussion.
Secondly, after the topic is determined, what I need to do immediately is not to think about how I should write, or to complain that I am depressed and don't know how to write at all. But through literature retrieval, we can see what research predecessors have done on this subject and how they have done it. Find the information related to the topic by searching the literature, and then sort out these materials. Sorting does not need to care about the conclusions and data details of the references, but to sort out the research purposes, research methods and analysis methods adopted by each document. Of course, you may not understand the analysis methods in the references at all, but it doesn't matter. You should first list all the analysis methods used in these references, such as linear regression, analysis of variance, mean t test, logistic regression, etc. , and lists the commonly used statistical methods in these documents. You need to be clear about the correspondence, that is, what kind of research purpose each analytical method is used to support and achieve, and what kind of conclusions can be drawn. You can achieve this step by reading the literature carefully.
Third, through the previous step, you should vaguely know the names of statistical methods commonly used in references related to your topic, and what purpose these statistical methods can help to achieve, or what conclusions can be drawn. At the same time, you will not be so afraid and confused about your topic, because maybe your topic has been done by predecessors and your paper has only been "copied" once. When I say replication, I mean repeating the previous research. In this case, you can conceive the topic by yourself, which is purely theoretical. You need to concretize your ideas, such as what you want to achieve, and naturally you need what data analysis methods you need. Of course, many papers will design a series of hypotheses to be verified in advance, which is also completed in this step, because there may be contradictory conclusions in the documents you find, and there may also be some research defects that you think (if you read too many documents, you will naturally have your own ideas), and put forward your own series of hypotheses, which can clearly guide the subsequent data collection and analysis.
Fourthly, the topics, assumptions and research methods can be determined through the first few steps, and then the specific research and data collection links can be considered. The most important thing in this link is to figure out what kind of data you should have and how to get it. In fact, it is very simple, because you have determined the statistical analysis methods, and each method has its own specific data type requirements, such as classified data (such as gender, nationality, grade, etc. ) and continuous data (such as age, height, weight, temperature, length, distance, etc. The simple and popular understanding of classified data is that these numbers are meaningless in themselves, and they are artificially given some meaning. There is no continuity between these data, and addition, subtraction, multiplication and division are meaningless, while continuous data are meaningful and can be used for some addition, subtraction, multiplication and division operations. By determining the type of data needed, we can roughly know the problems that should be paid attention to when collecting data. For example, in a questionnaire survey, it is usually clear how to design questions. Usually, two data types should be considered when designing questionnaires, because different option designs will lead to different data types. If you design a question with answer options of "Yes/No" and "Yes/No", then it belongs to classified data. If your answer option is "very satisfied-very dissatisfied" on the Richter scale, you can only count some percentages according to classified data, or score according to continuous data such as 12345, so that you can get the average. Therefore, it is very important to determine the data type in this step. If the data type is wrong, the collected data is completely useless.
Fifth, I won't elaborate on the specific data collection process. After data collection is completed, it is data entry. Remember to enter the original data, not the data after adding, subtracting, sorting and summarizing. A data input format is also required. Generally speaking, in the same situation, one line represents the data of a case or a questionnaire, and one column corresponds to a question in the questionnaire, that is, a variable. Therefore, after data input is completed, there are as many rows as there are indicators in the data.
Sixth, this step is what you should start to worry about. What if data analysis doesn't work? Because only here, the specific analysis process of data begins. I don't know what to do. I already know the analysis method. In this case, I can only find a textbook and then find the corresponding methods to introduce learning, or I really can't find anyone to guide me, find someone to help me and so on.
Finally. After the analysis is completed, start writing the whole paper.
PS: I also want to emphasize that there are some problems with college tutors now, because I have contacted so many students, and their point of view is "What if my statistical test results are not significant?" Doesn't that mean my research is meaningless? My assumptions are all wrong? " "My conclusion is inconsistent with the previous result. It seems that mine is wrong again." These two views are obviously wrong:
First, the source and object of data have changed, and the conclusions stipulated by WHO must be consistent with those of the predecessors;
2. Is Edison's first 999 failures in inventing the light bulb meaningless? Scientific research is a process of falsification, and it is falsified again and again to get close to the truth.
Third, if your assumptions must be correct, you don't need data validation. You can help the police solve the case, because you think your hypothesis must be correct, so just assume how simple it is to solve the case. But obviously, many tutors have not conveyed these correct views to students.