PS: The questionnaire analysis here only represents the questionnaire analysis with scale.
Because the Maturity Scale has often experienced the test of reality, the probability of its reliability and validity reaching the standard is relatively high.
However, the self-designed scale is prone to terrible results in reliability and validity tests.
In this way, we can easily carry out reliability analysis, exploratory factor analysis and even confirmatory factor analysis.
In addition, if there are two problems, the KMO value of factor analysis must be equal to 0.5, and generally we want to get the minimum value of 0.6.
In order to get a better result of reliability and validity, in the text description part, each topic in the same dimension should be given some psychological hints as far as possible, or the description should be as close as possible, so that the topics in the dimension can have better relevance, so that the reliability and validity will not be too bad.
Generally, it is enough to reach 0.6, and more than 0.7 is better. It is best to find the reliability of each dimension, and then find one as a whole. This is generally not difficult and easy to pass. If your data is not reliable, then proceed to the next analysis and delete the samples with no distinction between high and low scores.
Only KMO value and bartlett sphere test value of the scale were found. This may be the minimum requirement of validity test. Unless the instructor agrees, it is best not to use only these two values.
Most students can use it, and it is not easy to pass the analysis.
The most common problem is that the hypothetical dimension of the topic is inconsistent with the actual result.
In this case, it is generally handled as follows:
Only a few questions don't match.
Either delete it directly or keep it temporarily.
(2) most of the questions don't match.
Redesign the scale, collect data again and start over.
It is not recommended unless there are special circumstances. Because through exploratory factor analysis and confirmatory factor analysis, it is almost impossible for the actually collected questionnaire data to reach the ideal value.
If you see someone using this test method in their paper, and the indicators are beautiful, I can tell you responsibly that the high probability is that the data has been changed.
Special case 1: AMOS structural equation is used in the model verification stage, and the tutor requires confirmatory factor analysis in the validity test stage.
Special case 2: In the model verification stage, the AMOS structural equation is not used, and the tutor also requires confirmatory factor analysis to test the validity. (instructor stupid x)
In particular, the significant P value represents whether they are related or not, and Pearson or Spearman coefficient represents the degree of correlation.
Pearson or Spearman coefficient is meaningful only if it passes the significance test. The greater the absolute value, the greater the correlation, and the positive and negative represent positive and negative correlation.
The significance passes, but the coefficient is too small, so the correlation is also significant, but there is a significant weak correlation between the two, rather than representing an irrelevant small coefficient.
This is probably the simplest model. Just put the independent variables and dependent variables in and run them directly.
① Do you want to put control variables?
This is optional.
If you put control variables, try to put some hierarchical variables instead of multi-classification variables.
Hierarchical variables such as education (junior high school, senior high school, university, master's degree)
Multi-categorical variables, such as occupation
The assignment of hierarchical variables should correspond to its items as much as possible.
If you put a multi-category variable, try to delete it. If you want to keep it, you'd better make it a virtual variable
② Standard coefficient or standardized coefficient?
Standardization coefficient.
③ Do you want to do the VIF*** linear test?
Don't do it unless the instructor asks.
How big is the square of R?
There is no very strict standard for this indicator, which has a very profound impact on the values of tutors.
For the data collected in reality, I personally think that it is generally better to be greater than 0.2.
But I have encountered a situation that is greater than 0. 1, and my tutor also thinks it is acceptable.
This is a matter of opinion.
From a scientific point of view, it should be closely related to the scene you are studying.
However, the mediation effect model is easier to pass and less evasive to explain than the regulation effect model.
So, if you don't want to dig a hole for yourself, use the mediation effect model.
A method for quickly verifying the mediation effect model (quickly determining whether there is mediation, using informally)
Condition 1, intermediate variable, independent variable, dependent variable, the correlation is significant.
Condition 2, the regression model of independent variables and intermediate variables about dependent variables, the coefficient of intermediate variables is significant.
Satisfying the above two conditions, the mediation effect must be remarkable; If the independent variable in condition 2 is also significant, it is a partial mediation effect; If it is not significant, it is a complete intermediary effect.
In other rare cases, Suo Beier is used to test the mediating effect.
If amos is not asked by the tutor to verify the mediation effect, we can try to use spss regression to test the mediation effect.
A method to quickly verify the conditioned effect model (quickly determine whether there is a regulatory effect, use informally)
First, calculate the adjustment coefficient (standardized independent variables and intermediate variables can be multiplied)
The regression model of independent variables, regulatory variables and regulatory factors on dependent variables shows that the coefficient of regulatory factors is significant.
WeChat official account: alone5400