A scale is a measurement tool. There are many scale design standards, and the design of each scale has its own characteristics. The characteristics of different scales also determine the measurement scale. The scale commonly used in data analysis is Li Ke special scale. The Likert scale was formed in 1932 by Likert, an American social psychologist, on the basis of the original summation scale at that time. It is a method of market research. The table generally sets five answers. It is recorded as 1, 2, 3, 4, 5, and there are also seven scales.
Because the Likert scale is more commonly used, this article uses the analysis of the Likert scale ( hereinafter referred to as scale data ) for illustration. Generally, it can be divided into six parts. The first part is the description of the characteristics of the sample data, the second part is the reliability of the scale, the third part is the difference relationship, the fourth part is the influence relationship, and other descriptions. Next, we will explain them one by one.
- Data characteristics
After the sample data is collected, first describe the data characteristics, the purpose is to have a deeper understanding of the basic situation of the sample, such as statistical descriptive analysis or frequency analysis, to further understand the characteristic behavior or basic attitude of the sample population. - Scale reliability
Scale reliability can be described through reliability and validity analysis, what is reliability? Reliability is simply to study whether the sample data is true and credible. What is validity? Validity refers to whether the design of the measurement items of the questionnaire is reasonable or not, which will be explained in detail below.
Generally, after conducting scale reliability research, if researchers want to continue mining sample information and get more effective conclusions, the common ones are difference relationship, correlation and influence relationship. First, explain the difference relationship . Whether there is a difference in product satisfaction by gender, and whether there is a difference in satisfaction with shopping mall service attitudes with different educational backgrounds, usually use variance analysis, t-test or chi-square test for research, often use t-test or variance analysis, the difference between the three The data type is different, which will be described in detail below.- Correlation and influence relationship
In addition to studying the difference relationship, you can also study the correlation and influence relationship. The correlation relationship studies the relationship between two variables. Through the correlation relationship, researchers can roughly understand the basic relationship between variables and whether there is a correlation relationship. Most of the research on correlation is to pave the way for regression analysis, because a correlation does not necessarily have an influence relationship, but an influence relationship must have a correlation, so generally researchers conduct correlation research before performing regression analysis. - other
Sometimes multiple samples are studied for classification (cluster analysis) or weight calculation of indicators, etc. It will be explained next.
1. Data characteristics
Generally, data characteristics are described using frequency analysis or descriptive analysis to understand data distribution characteristics, and sometimes visual graphics can also be used to describe the results more intuitively. Case background: The questionnaire mainly studies employee satisfaction. Employee satisfaction is divided into four variables: personal development, job characteristics, leadership management, and job returns. The four dimensions are represented by 2 to 4 scale questions ( see Questionnaire data can be pasted into a browser for use ). For example, if you want to study the respondents’ “whether they agree with the point of view that labor pay and work income match”, the results of the analysis are as follows:
From the table results, we can get a total of 389 valid samples collected, among which the largest proportion of people who agree with the above views is 164 people, and the proportion of people who hold "agree" and "completely agree" views is 42.16%+12.08%=54.24 %, so it shows that more than half of the surveyors agree with the view that “labor pay and work income match”. At the same time, the scale questions can also be cross-analyzed with demographics, such as analyzing the views or opinions of different genders and different occupations. The next step is to check the reliability of the scale.
2. Scale reliability
The reliability of the scale can be explained from the perspective of reliability and validity. Reliability analysis means whether the sample really answered the question. Usually, reliability analysis can only analyze scale items. Reliability analysis is only for quantitative data.
Validity analysis refers to whether the design of the research measurement items is reasonable. Under normal circumstances, validity analysis is only for scale data, and non-scale questions such as multiple-choice, single-choice gender and other topics cannot be analyzed for validity. If you want to analyze the validity, it is recommended to use "content validity", that is, to describe the process of questionnaire design in detail in words, to describe clearly what the questionnaire does, what it is useful for, and why it is reasonable, and it is certified by experts. This shows that the questionnaire design is reasonable and effective. Generally speaking, there must be reliability before validity, so the reliability analysis is carried out first.
Reliability is divided into intrinsic reliability and extrinsic reliability. For specific differences, please refer to the recommended article (provided below). There are many calculations for the reliability coefficient of intrinsic reliability, but the commonly used one is the Cronbach coefficient . So the example It is also described in terms of Cronbach's coefficient. Because this questionnaire contains a total of 4 dimensions, the results of reliability analysis by dimension are organized as follows:
Next, check the validity of the questionnaire. Since the questionnaire scale has four dimensions, the number of dimensions is set to 4. The operation is as follows:
The result is as follows:
From the above table, we can see that the 12 items of the validity analysis scale are divided into 4 dimensions. From the results, the structure is good, and it can be seen from the above table that the KMO value is 0.916>0.6, and the p value is less than 0.05. , the cumulative variance explanation rate value is 74.78%, indicating that the four dimensions can extract most of the item information. Therefore, it shows that the research data has a good level of construct validity. More content can also refer to:
3. Difference relationship
Difference relationships are generally used to compare the differences between two or more sets of data. Common methods for the study of differences are t-test, variance and chi-square test. The difference between the three is that the data types are different. The t test (here the value of the independent sample t test) and the variance (here refers to the one-way analysis of variance) require the data independent variable to be a fixed variable, the dependent variable is a quantitative variable, and the chi-square test requires the independent variable Both the dependent variable and the dependent variable are definite variables. For t-test and variance, if the independent variable is two groups such as male and female, the t-test is generally used. If the independent variable is more than two groups such as "primary school", "junior high school" and "high school" , generally using analysis of variance. For example, if you want to study "whether there are differences in the personal development of different positions" in the questionnaire, since the group of positions is more than two groups, use analysis of variance (one-way analysis of variance) to describe, first of all, the four scales under personal development Combining the title into a variable.
Proceed as follows:
- First click [Generate Variable] in [Data Processing];
- Then choose the four scale questions under Personal Development
- Because the scale questions are combined into one dimension, the average value is used for processing
- Finally, name the dimension and click Confirm to process
Then, a one-way analysis of variance will be performed with position as the independent variable and personal development as the dependent variable. The results are as follows:
From the results, it can be obtained that the F value of the model is 3.061, and the p value is 0.028 less than 0.05, so the model is significant, indicating that there are differences in the personal development of different positions. Similarly, independent sample t-test can also be used to analyze whether there are differences in the personal development of different genders. The results are as follows:
From the results, it is also found that the t value is -2.597, and the p value is about 0.01<0.05, so it shows that people of different genders have differences in personal development. For more content, please refer to:
4. Correlation and influence relationship
To deal with research differences, scale questions can also study correlation and influence relationships. For example, if you want to study the impact of personal development, job characteristics, and leadership management on work returns, first combine multiple scale questions into one dimension, and the steps are the same as above. , and then analyzed, the results are as follows:
Because the research on personal development, work characteristics, and the relationship between leadership and management on work returns, the data are all quantitative variables, so linear regression is used for analysis. First, the correlation test of the analysis items is carried out before the analysis. The results are as follows:
As can be seen from the above table, using correlation analysis to study the correlation between work rewards and leadership management, job characteristics, and personal development, the results show that there are correlations between work rewards and the three, so the next step is to regression analysis:
First look at the F test, the results are as follows:
From the above table, it can be concluded that the regression sum of squares dimension is 149.909, the residual sum of squares is 144.613, the total sum of squares is 149.909+144.613=294.522, the F statistic is 133.033 and the p value is less than 0.05, so it shows that the model construction is meaningful, at least one self A variable has an effect on the dependent variable. View the regression results as follows:
A linear regression equation was established with work return as the dependent variable and leadership management, work characteristics, and personal development as independent variables. From the results, the p values of the three variables of leadership management, work characteristics, and personal development were all less than 0.05, and the regression coefficients were all If it is greater than 0, leadership management, job characteristics, and personal development all have a positive impact on job returns. Because I only want to study the impact relationship, I won’t go into details about the model of the regression equation.