1. Teaching content
9Does your paper have such problems or errors?
9.1 Introduction
Psychological empirical papers need to involve statistical analysis of data, and some errors are prone to occur in this process. It mainly involves sampling, selection of research variables, reliability and statistical analysis.
9.2 Analysis
(1) Sampling
① Use of biased samples. Many studies are for the convenience of illustration and use convenient sampling. Such samples are obviously biased and do not conform to the principle of randomness.
② Failure to pay attention to the deviation caused by the subject's response tendency. By means of online surveys, telephone surveys, letter surveys, etc., subjects can decide whether to respond to the survey. Obviously, there is a deviation between the subjects who actively respond and those who do not respond.
(2) Research variables
Only consider the explicit variables, without considering or studying latent variables (traits or abstract psychological factors). For example, when studying the influence of teaching methods on test scores, there are obviously many factors that affect test scores. Some are very important, such as achievement motivation. If we do not study these, our conclusions will be greatly reduced.
(3) Reliability of internal consistency of the scale
We often use α internal consistency coefficient to measure the reliability of a scale, that is, high reliability. However, internal consistency reliability is suitable only when the various items of the scale have some internal correlations and reflect common things. If a scale is composed of some scattered subscales, and the total score has no practical significance, the internal consistency reliability is also high, but in fact there is not much correlation between the subscales. In addition, when some irrelevant questions are piled up to form a questionnaire, the internal consistency reliability is relatively high, but obviously there is no substantial internal connection. The high alpha coefficient does not mean that the scale measures a single-dimensional construct.
(4) Statistical analysis
① For single-factor multi-level or multi-factor design, use t-test to analyze, and use analysis of variance if it is correct.
② When the analysis of variance found that the interaction effect was significant, the simple effect was not further analyzed.
③ In correlation analysis, a relatively high level of significance is used to illustrate the close relationship between variables. For example, the correlation coefficient is only 0.12 and the significance level is p<0.001. In fact, the relationship between the two is very weak. And even if the correlation coefficient is high, it does not mean that the two are highly correlated, because this high correlation may come from the common influence of some common factors.
④ In the regression analysis, the significance level of the independent variables in the regression analysis table is used to illustrate the great influence. In fact, we should look at the measurement coefficient, β value, etc.
⑤ In the discussion section, too many specific data results of other people's researches are cited, even including the significance level. In fact, it is not necessary to
quote the main conclusions.
2. Remarks
Need a full set of information or add QQ1564658423.