Dear all
I am assigned a task to perform recurrence analysis on our data.
The data we have are robot repairs. We want to see repair events and cost in response to accumulated transactions using recurrence analysis.
I attached a simulated data for this question. It only has three bots since we haven't decided how to arrange the data and I just put those three bots manually. I will update with more simulated data once we have a clue on how to format the data for the analysis.
Explanation of the simulated data: (You can skip this long description to the question if you want)
The frist two columns are 'BOT ID' and 'Site'. Working sites for bots can change as the transaction increases. In the example data, bot 9880 changed site several times. After some study and discussion with great JMP community members, we conclude that Site is Time-varying effect. One concern is that JMP does not have an automatic function for calculating MCF for time-varying effects. I will update the post with more simulated data later, so you can determined how to perform the calculations.
'Transaction Date' and 'Accumulated Transactions' is the time and amount of the transaction when the repair occurred. We want to use transactions in the analysis, because the bot can sit in the sites for a long time without a transaction due to different reasons.
'Censor' columns has row values 0 and 1. Each bot has a 0 censor value at its last observed transaction and this is the closing for the bot. In addition, if a bot changes site, a row with the initial transaction in the new site with 0 censor is added to indicate the start of a new site.
'BotSeries' has two values, indicating two types.
'Repair Date' is when the repair occurred.
'Usage' has two values, SOW (standard), L(loaner). Loaners probably change site as they are loaned to some sites in need.
'Part' and 'Price' are the parts repaired or replaced, and the price of the part.
'Service Center' is where the bot is repaired.
Questions:
Service Center is a new variable that we want to study. The logic to fill this column is that if a repair is easy to do, the repair is performed in the same site where the bot works. If the repair is complicated and beyond site capacity, the bot will be shipped to either PLC or ITC (two centers in different states in the US). The choice between PLC and ITC is largely affected the location of the site. For instance Site B will go to PLC. Other sites go to ITC starting around last year. Before that, there was only one ITC service center.
I don't know what effect is Service Center now. Is it a time-varying effect? It is highly affected by Site, which is time varying. But it is not where the bots primarily located like Site. Or it can be used as a repair mode, like Parts can be interpreted as a failure mode? So we can specify Service Center to the Cause column to generate recurrence analysis for each Service Center value? You will see what I mean by Cause column in the screenshot.
One more question is that BOT9880’s last two rows. The bot moved to a new site, did not have any repair after some transactions and was censored at the time of the study. So I have two rows with Censor 0, the first indicates the start of a new site. The second indicates the closing of the bot. Is that the right way to do this?
Thanks a lot.