I conducted a study in which i have daily attendance data (Time In and Time Out) for 120 individuals.
Specifically, attendance data are for every first M/W/F of each month, although i can always code for more (i.e., 3*12= 36 time points per person). For each such time point/day, i have Time In and Time Out as continuous variables (clock minute time), and i also have dummy codes for come early, come on time, come late, leave early, leave on time, leave late-all for each day. Finally, I also have variables measured once (e.g., personality) as well as three times (e.g., Job satisfaction).
Ideally, i would like to examine:
1)If employees can be grouped in meaningful attendance categories based on their daily attendance behavior throughout the course of the year.
2)Whether such categories vary by the day/month measured or are relatively stable.
3) Whether such categories can be predicted by variables (e.g., personality), even by taking into account their possible variation due to day/month.
4) Whether such categories coincide with levels of job satisfaction, that is go up/down together.
After much research, i still am not sure the type of analyses i need to conduct. However, i am very motivated to learn.
Thank you very much for your time and for reading this. Any help would be greatly appreciated!
Would it be a regular latent class analysis then? What about the fact that my indicators (dummy codes) are repeated measures? There would have to be a command in the syntax that accounts for this, correct?
Is there an example that you can direct me to on the website?
So i wouldn't be able to model the 36 dummy-coded observations per individual, and form latent classes based on these within person observations? (N=36 observations per person X 120 employees)
Rather, i would have to get a single dummy code observation per employee(an average over certain days as you indicated-can i do that with dummy codes?) and form latent classes from the between person observations (N=120 employees)?
I hope the above make sense, i tried to describe what is in my mind. I appreciate your time and help in this!
There are many ways to do this. I can't make that decision. One suggestion is to average over some of the days. You don't need to do that. For further guidance, you should seek the help of a statistical consultant. We can't provide that level of help on Mplus Discussion.
I understand and thank you for your help and time in this!
June H. Kim posted on Wednesday, November 30, 2016 - 8:29 am
My apologies, I am new to this to board and wasn't sure where to post my query. I am running a LCA on state policy characteristics. The unit of analysis is state/year, with a total of 784 observations (49 states*16 years). I used the cluster option to let MPLUS know that there are repeated measures by state. However, I am concerned that I am not accounting for this properly, as the few examples of RMLCA I have seen indicate the need to set the data in wide form (currently, the data is in long form with 16 rows for each state). Do you think this is appropriate? I can send license number or any other info that may help.
variable: names are fipscode year state pmp2012 a_law a_hhs a_pla a_other pr_law pr_pla pr_pd research man_reg access_p access_d access_r access_other pres_access disp_access law_access pdmpout disreport drug_4 must_access share_pim;