Many students collect questionnaires to obtain the data needed for thesis research, but how should the collected questionnaires be analyzed? Questionnaires can generally be divided into two categories: non-scale and scale questionnaires. Different types of questionnaires have different analysis ideas. Today I will discuss with you what the general analysis ideas are after getting a questionnaire.
1. Questionnaire analysis ideas
SPSSAU provides the following five types of scale data analysis ideas for reference. They are: Research on the influence of scale questionnaires, research on the mediation effect and moderating effect of scale questionnaires, research on the weight of scale questionnaires, research on the differences of "quasi-experimental" questionnaires, and clustered samples Questionnaire-like study:
1. Research on the influence of scale questionnaires
Influence relationship research is very common, and its core issue is to explore the mutual influence and relationship between different variables. Correlation analysis can usually be used to study the relationship between variables, such as whether there is a relationship between variables, how close the relationship is, etc., and then use regression models to study the regression impact relationship between variables.
The research and analysis ideas of the influence relationship of scale questionnaires are as follows:
1.1 Sample background analysis
First conduct basic frequency analysis on the collected data, such as statistics on gender, age, monthly income level, and occupation distribution. The results of SPSSAU frequency analysis are as follows:
The analysis in this part should first describe the sample size, and then describe the sample background information respectively, especially the important information points.
For example: There were 299 questionnaires in this survey, of which 71.237% were women; the highest proportion was 45.667% of people under 20 years old; the monthly income was less than 2,000 yuan The highest proportion of the population is 36.667%; the highest proportion of students is 46.667%.
1.2 Sample characteristics and behavior analysis
After statistics and description of the background information of the research sample, it is necessaryto further analyze the basic attitudes, characteristics, behaviors, etc. of the research sample.
For example, analyze the questions in the questionnaire related to the basic attitude of the sample: "Why do you learn a foreign language?", "How long do you have experience in learning languages online?", "How many courses have you purchased?". These three questions are all single-choice questions, so the frequency of selections is calculated separately (refer to the figure below). When reporting,you should first focus on the options with a higher selection ratio and highlight the key points.
1.3 Indicator classification analysis—exploratory factor analysis
When studying the influence relationship, the questionnaire usually involves a lot of scale questions.If it is not completely sure how many dimensions the scale questions should be divided into, you can Use factor analysis to condense, obtain several dimensions (factors), and find the correspondence between the dimensions and the items.
1.4 Reliability analysis
The purpose of reliability analysis is to study whether the sample data is true and reliable. Generally speaking, it is to study whether the respondents answered each question truthfully. If the respondent does not answer truthfully, the reliability is not up to standard. Reliability is only studied for scale questions, and cannot be analyzed for background information items such as gender and age. Reliability can be divided into the following five categories, among which Cronbach's alpha reliability coefficient is the most commonly used:
1.5 Validity analysis—exploratory factor analysis, confirmatory factor analysis
The purpose of validity analysis is to determine whether the research questions can effectively measure the variables that researchers need to measure. In layman's terms, it means whether the measurement questionnaire questions are accurate and valid. When the reliability analysis fails to meet the standards, the validity analysis will inevitably fail to meet the standards as well. Validity was studied only for scale questions. Validity can be divided into the following 4 categories:
1.6 Research variable descriptive analysis
The purpose of research variable descriptive analysis is to study the overall attitude of the sample towards the variable. When studying variable description analysis, reverse questions need to be processed in reverse (the customary processing method is that the larger the score, the more satisfactory). For example, when a score of 1 represents "strongly agree" and a score of 5 represents "strongly disagree," it needs to be reversed to a score of 5, which represents "strongly agree," and a score of 1, which represents "strongly disagree." Analyze by calculating the mean or median of variables, or use a line chart to display the ordering of the mean values of variables, etc. .
The SPSSAU description analysis results are as follows:
1.7 Variable correlation analysis - correlation analysisStudy the relationship between variables through correlation analysis, includingwhether there is a relationship and a close relationship Degree. Usually a variable is represented by multiple questions, so before performing correlation analysis, it is necessary to calculate the average of multiple questions to represent the corresponding variable (SPSSAU generates variable-average values for processing).
1.8 Research hypothesis testing analysis - regression analysis
On the premise that the data is relevant, it is meaningful to study the regression influence relationship. Therefore, regression analysis needs to be placed after correlation analysis, and regression analysis is usually used to verify hypotheses. If the dependent variable is quantitative data, then linear regression analysis or SEM structural equation model can be used for hypothesis verification; if the dependent variable is categorical data, then logistic regression analysis can be used for hypothesis verification.
1.9 Difference analysis - analysis of variance, t test, chi-square analysis The purpose of difference analysis is to unearth more valuable research conclusions, such as the research on the relationship between male and female samples Whether the variables have differential attitudes. There are usually three analysis methods for difference analysis, namely analysis of variance, t test and chi-square analysis. Scale questionnaires usually use variance analysis and t-test, while non-scale questionnaires generally use chi-square analysis.
2. Research on the mediating and moderating effects of scale questionnaires
In scale questionnaire research, the study of mediating effects and moderating effects is also relatively common.The research of mediating effects and moderating effects is an extension of the study of influence relationships. These two types of research are commonly used For academic research, the rest part is basically similar to "influence relationship research".
The research ideas on the mediating and moderating effects of scale questionnaires are as follows:
The overlapping parts will not be described again, and the additional new parts will be explained as follows.
2.8.1 Mediating effect
The mediating effect is when studying the influence of Such a relationship; for example, job satisfaction (X) will affect innovation atmosphere (M), and then affect final job performance (Y).
2.8.2 Moderating effect
The regulating effect is whether it will be interfered by the regulating variable Z when studying the impact of , this influence relationship is interfered by whether you drink alcohol (Z), that is, whether the influence amplitude when drinking is significantly different from the influence amplitude when not drinking.
3. Research on the weight of scale questionnaires
The focus of weight research on scale questionnaires is the weight score of each indicator rather than the impact relationship. By calculating the weight of each indicator or question Weight score, build a complete weight system, and put forward scientific suggestions based on the weight of each indicator.
The research and analysis ideas for the weight of scale questionnaires are as follows:
3.6 Research on weighting system
The weight of an indicator refers to the quantitative value of the relative importance and value of indicators at all levels in the entire evaluation system. The weight value of each indicator will be recorded as a decimal between 0 and 1, with 1 as the entire indicator. The sum of the weights of the system.
4. Research on differences in “quasi-experimental” questionnaires
"Quasi-experimental" questionnaires refer to questionnaires with an experimental background. "Quasi-experimental" questionnaires usually take the study of differential relationships as the core content, and are generally conducted using single-factor analysis of variance, multi-factor analysis of variance, t-test and other methods. Research.
Analysis ideas for research on differences in “quasi-experimental” questionnaires:
4.5 Interaction studies
Interaction research refers tostudy the impact of multiple categorical independent variables X on dependent variable Y (Y is quantitative data), that is, research The difference in the magnitude of the influence of multiple categorical independent variables X on Y when they are at different levels. For example, we now want to study whether there are differences in yield between different fertilization methods and different rice varieties, and whether the interaction between fertilization methods and varieties has an impact on rice yield. Studies similar to those described above are interaction studies.
5. Cluster sample questionnaire research
When studying clustered samples, the first thing that comes to mind is the "classification" of the samples, that is, the sample population should be divided into several categories; after the categories are divided, it is usually necessary to compare the differences between different categories of people, such as different Differences in attitudes and behaviors between category groups, etc.
The analysis ideas for cluster sample questionnaire research are as follows:
5.6 Cluster analysis
Cluster analysis can perform cluster analysis on samples (Q-type clustering) or cluster analysis on variables (R-type clustering). Cluster analysis is classified as follows:
5.7 Verification of clustering effect
Verification of clustering effect is different from other analysis methods. Other analysis methods can be tested through p-value. Verification of clustering effect requires certain research experience and comprehensive judgment combined with professional knowledge. Good clustering effect can effectively identify sample characteristics. Comparison of feature differences of clustered samples is usually performed using variance analysis. Sometimes the clustering effect can also be judged through discriminant analysis.
2. Ideas for analysis of non-scale questionnaires
Under normal circumstances, non-scale questionnaires are used to analyze the current situation of a certain topic, understand the basic attitude of the sample, study the current situation or attitude differences of different types of samples, and then provide meaningful suggestions and measures based on the analysis conclusions.
The framework of non-scale questionnaire analysis is shown in the figure below:
The parts that overlap with the analysis ideas of the scale questionnaire will not be described in detail, and the new parts will be supplemented.
3. Basic status analysis
Fully understand the current situation of the sample and the attitude of the sample. Combined with the results, you can analyze the attitude differences and current situation differences of different groups, or further study the influence relationship.
When conducting research, you should not stick to the use of analysis methods. This part will use simple and easy-to-understand frequency and percentage descriptions. It is best to combine Various graphic displays, such as bar charts for multiple-choice questions and column charts for single-choice questions.
Example:For example, when analyzing the multiple-choice question "Factors affecting the purchase of courses", the SPSSAU multiple-choice question analysis results are as follows:
During text analysis,researchers need to pay more attention to options with a higher selection ratio. It can be seen from the analysis results that the selection proportions of "teaching quality" and "course content" are significantly higher than other items, while the selection proportions of "discounts" and "others" are relatively small.
4. Sample attitude analysis
If the questionnaire involves questions related to the cognitive attitude of the sample, frequency analysis or multiple-choice question analysis can be used to summarize and further understand the characteristics of the sample (refer to the above process for analysis).
5. Difference analysis
After laying the groundwork in the previous section, it's time to compare the differences. You can analyze the differences in attitudes of different sample groups on the items, or conduct comparative analysis on the differences between different groups of people on the basic status quo items.
In terms of research methods, for research on the relationship between non-scale items, that is, the relationship between classification and categorical data, Chi-square analysis< should be used /span>.
Example:Research why learning English is different in different professions? The results of SPSSAU chi-square analysis are as follows:
From the results of the chi-square analysis, we can know that there are significant differences in the reasons for learning English among different professional groups (chi=114.089, p=0.000<0.01). The specific differences can be analyzed by comparing the percentages in brackets, or by viewing the stacked bar chart below. Intuitive comparison.
From the above figure, we can intuitively see that students study foreign languages mainly for exams (46.43%), company employees mainly study foreign languages to improve their work skills (38.71%), etc., which will not be elaborated here.
6. Impact relationship analysis
Next, you can study the influence of certain factors on the attitude of the sample. When the dependent variable Y is categorical data, logistic regression analysis should be used to study the influence relationship. Logistic regression analysis has the following three categories, which are explained as follows:
Example:You want to study what factors affect the dependent variable: "Would you be willing to share your course with others?" At this time, the dependent variable is a binary variable, so binary logistic regression analysis should be used for research.
7. Others
If the questionnaire contains data stored in quantitative variables, such as height, weight, etc., relevant analysis and research can be performed. Or use analysis of variance or t test for difference analysis, etc.
- The core of non-scale questionnaire analysis ideas
The core of this type of research framework lies in "grouping".
- The first thing is "grouping", that is, grouping each question. For example, there are 30 questions in the questionnaire. How many aspects can these 30 questions be divided into? For example, five aspects include basic background, cognition, attitude, behavior, and reasons.
- The second thing is to analyze the "grouping" as a part. For example, for the basic background of the sample mentioned above, frequency analysis can be used to statistically analyze the data.
- The third thing is to cross between grouped items and grouped items. For example, the basic background is different from "cognition", "attitude", "behavior" and "reason". Usually a crossover analysis is used.