What if the reliability coefficient is low? How to calculate the reliability coefficient?

One 、 Research scenarios

Reliability analysis is used to measure whether the sample answer is reliable , That is, whether the sample actually answers the questions of the scale （ Important note ： Reliability analysis is only for Scale data , Generally, reliability analysis is not carried out for non scale data ）; Reliability analysis is only for quantitative data .

Two 、SPSSAU operation

1. operation

SPSSAU Left instrument cluster “ Questionnaire research ”→“ reliability ”;

2. Calculation of reliability coefficient index

（1）Cronbach α coefficient

among N To measure the number （ Put in SPSSAU Number of analysis items ）,  Represents the total variation after data summation ,  It means the first one i Data variation of item ,   Represents the sum of variation of various data . It can be seen from the formula that , The number of measurement items will be right Cronbach α Influence relationship of reliability coefficient , When the number of analysis items is more ,Cronbach α The reliability coefficient may be higher . The minimum number of measurement items is 2 individual , At this time, the reliability coefficient may be relatively lowest .

It can be seen from the formula that , The number of measurement items will be right Cronbach α Influence relationship of reliability coefficient , When the number of analysis items is more ,Cronbach α The reliability coefficient may be higher . The minimum number of measurement items is 2 individual , At this time, the reliability coefficient may be relatively lowest .

（2） Half coefficient

The half coefficient involves Spearman-Brown Sum of coefficients Guttman Split-Half coefficient . among Spearman-Brown The coefficient is divided into equal length and unequal length . They are described as follows ：

If it is equal length , Equal length at this time Spearman-Brown The coefficient calculation formula is as above , among R Represents data split into two parts （ First split the data into two parts , Then sum them separately , Get two columns of data respectively ） Correlation coefficient value of . If it's not the same length , That is, the number of analysis items split into two parts is inconsistent （ That is, odd term ）, At this time, the length is not equal Spearman-Brown The coefficient is calculated as follows ：

In the above formula R Is the correlation coefficient of two parts of data ,k1 and k2 Respectively represent the number of analysis items of the two parts of data ,k=k1+k2.

meanwhile ,SPSSAU And to provide Guttman Split-Half coefficient , It can also be used to measure reliability . The calculation formula is as follows ： In the above formula ,  Represents the variance of the sum part of the whole ;  and    Each represents the 1 part , The first 2 Partial variance .

（3）McDonald Omega

McDonald's ω The calculation principle of reliability coefficient is to use factor analysis to condense information , Then get loading Load factor value , And then calculate . The calculation formula is as follows ：

In the above formula loading Is the load factor value ,uniqueness by 1-loading^2. As can be seen from the above formula ,loading When the absolute value of the whole value is larger ,McDonald's ω The higher the value of the reliability coefficient .

（4）theta coefficient

In the above formula N Is the number of analysis items ,   Is the maximum eigenvalue . You can see from the above formula , When the number of analysis items is more ,theta The reliability coefficient is likely to be larger , And the larger the largest characteristic root ,theta The greater the value of the reliability coefficient .

3、 ... and 、SPSSAU analysis

（1）Cronbach Reliability analysis

Additional explanation ： Standardization Cronbach α Coefficient values and Cronbach α The function of coefficient value is the same , The criteria are exactly the same , Generally, it is rarely used , Use it directly Cronbach α The coefficient value is used to analyze the reliability .

It can be seen from the above table that ： The value of reliability coefficient is 0.934, Greater than 0.9, Therefore, it shows that the reliability quality of the research data is very high . in the light of “ Item deleted α coefficient ”, After any item is deleted , The reliability coefficient will not increase significantly , Therefore, explain that the item should not be deleted .
in the light of “CITC value ”, Analysis item CITC All values are greater than 0.4, It shows that there is a good correlation between the analysis items , It also shows that the level of reliability is good . in summary , The reliability coefficient of the research data is higher than 0.9, It shows that the reliability of data is high , It can be used for further analysis .

Cronbach α Specific calculation of reliability coefficient ：（ Because there are too many data in this case, it is inconvenient to show , So give an example ）

（1） 

（2） 

among N Is the number of analysis items ,n Is the sample size ;

（3）Cronbach α Reliability factor

（2）Cronbach Reliability analysis - Simplify the format

The analysis results come from SPSSAU

The simplified format only provides the number of items 、 Sample size and Cronbach α coefficient . It can be seen from it that , The number of items in this analysis is 14, The sample size is 200 And the reliability coefficient is 0.934.

Additional explanation ： The standard of reliability coefficient is as follows ：

Four 、 Suggestions for improving the reliability coefficient

Reliability analysis is not up to standard in any case ？

It is recommended to check according to the following seven steps ：

First of all ： Use ‘ Describe the analysis ’ Check whether there are strange outliers , If so, use ‘ Data processing -> outliers ’ Analyze after function processing ;

second ：‘ Non scale ’ Data cannot be analyzed for reliability , Just use words to describe and prove why the data is credible , For example, how to design questionnaires and collect data , Why is the data credible , Have you ever handled outliers ;

Third ： If the value of the reliability coefficient is still very low （ For example, below 0.5）, At this time, we can consider merging all the scale questions for a reliability analysis （ The more questions, the higher the reliability coefficient ）;

Fourth ： If there is a reverse question in the data , You need to use it first ‘ Data processing -> Data encoding ’ Deal with the reverse question and then analyze ;

The fifth ： Delete unreasonable items , Leave meaningful items ;

The sixth ： Increase the sample size , The larger the sample size is, the higher the reliability is usually ;

The seventh ： Try to design the questionnaire in one dimension 4~7 A question is better , The more questions, the higher the reliability , And if it does not meet the standard, individual unreasonable items can be deleted .

5、 ... and 、 To solve a problem

1. omega How to analyze the reliability coefficient ？

SPSSAU Provide McDonald's omega Reliability factor ,McDonald's ω Interpretation of coefficient and common Cronbach The coefficient interpretation is basically consistent , On the principle of McDonald's ω The coefficient is carried out by using the idea of dimensionality reduction and information concentration of factor analysis . In addition, if the analysis item is greater than 20, No output at this time ‘ Item deleted McDonald's ω coefficient ’.

2. Items deleted in reliability or validity analysis , Whether it needs to be retained in the future ？

If we have been analyzing the reliability （ Or validity , Or other analysis ） When you think that an analysis item is unreasonable, you need to delete it , Then all subsequent analysis methods generally need to be synchronized , It is not to delete the data directly , Instead, the item is not analyzed directly during analysis .

3. How to analyze the reliability of non scale ?

A scale is something similar to “ Very dissatisfied , dissatisfied , Very satisfied with ” Such questions , The first 1 Options use 1 Fractional representation , The first 2 Options use 2 Fractional representation , Go down in turn , The higher the score, the more satisfied , Or the more dissatisfied . Only scale questions can be used for reliability analysis （ And validity analysis ）.

Reliability analysis is a research method to study whether the data quality is reliable , If it is a non scale question , There is no way to use research methods for analysis , Only words can be used to describe , To prove that the data quality is reliable , It is suggested to explain according to the following points .

First of all ： Describe the process of data collection in words , For example, paper collection , Network collection, etc ;

second ： Describe the validity of the sample in words , For example, whether the collected samples meet the research needs , The sample population collected is an effective sample population , For example, study the online shopping behavior of college students , The sample population must be college students ;

Third ： Describe the processing of invalid samples in the data in words , For example, use SPSSAU Invalid sample function of , Will choose the same answer more than 70% Set the sample of to invalid sample, etc ;

Fourth ： Other descriptions that can be used to demonstrate that the data quality is guaranteed .

6、 ... and 、 summary

Usually , Reliability analysis uses α The coefficient indicates the reliability quality of the scale , That is, the reliability of the sample answer . Reliability analysis and prediction test will pay more attention to the quality of the scale , That is, whether the reliability quality is not up to standard due to the problems in the design of the scale questions , If something goes wrong , Then the question method needs to be modified , Or delete the question . Formal reliability analysis only needs to focus on a coefficient , Usually , This value is greater than 0.7 that will do , Sometimes the standard can be relaxed to 0.6.

Please log in for more dry goods SPSSAU Official website ：