SPSS Statistics

Expand all | Collapse all

Data Collection with Reliability

  • 1.  Data Collection with Reliability

    Posted Wed August 04, 2021 08:34 AM
    I am contacting with an issue regarding running reliability tests for my dataset.
    It seems that for three of my scales, the reliabilities are very low (i.e., less than .6). I haven't been able to find the reason for why the reliabilities are low (I've already checked the recoding and everything was done correctly). For one scale in particular, the TRIM-18, has an alpha of .637 (after recoding), but .91 (TRIM-18 Original; before recoding). I'm not sure what had gone awry to cause the alpha to change.
    The last scale, which contains dichotomous choices (i.e., true or false), also has very low reliability. I understand that the Kuder-Richardson formula is needed to measure reliability for dichotomous choices, but was told that the way to calculate the reliability is the same as calculating for Cronbach's alpha. Can you please confirm if this is correct?
    To conclude, I'm just confused as to why these scales have such low reliabiltities. I have attached the SPSS output file to this email of the three scales I am writing about.
    I am not sure if this is the correct email to contact, but if it is not, then I would appreciate it if you could kindly redirect me to someone who can provide help for this issue.
    Thank you for any help you can provide.


    ------------------------------
    Linghsi Liu
    ------------------------------


  • 2.  RE: Data Collection with Reliability

    Posted Thu August 26, 2021 04:22 AM
    Hi, Linghsi, 
    You have raised two queries, I am replying one by one. By the way, could not find any file that you have mentioned, so answering you generally:

    Query 1: Possible reasons of low Cronbach's Alpha
    Reason 1: Inter-correlations among the variables are low. Check you correlation matrix. If correlations among the variables are high (higher than 0.5), there should not be a problem. 
    Reason 2: No. of variables are low. Say, if you are testing for just 2 variables, the Cronbach's Alpha may turn out to be low. In this case, if no. of variables cannot be increased, then low (0.637) Cronbach's Alpha is good enough. 
    Reason 3: Grouping / Factor Analysis of the variables is not done properly. Please follow the factor analysis steps properly, remove lower correlated variable, group only the high correlated variables and then check for Cronbach's Alpha of the high correlated variables only. In this case, all your further analyses will have to have only the variables that you have kept (high correlated variables).

    Query 2: Cronbach's Alpha for binary variables
    I think you can use Cronbach's Alpha for binary variables also.

    ------------------------------
    Vandit Desai
    ------------------------------