Simple, possibly stupid question: If I want to have a linear regression of a single tag e.g. BA:TEMP.1 against time, how do I do this?
Solved
Regression of a single tag against time
+2Best answer by Jef Vanlaer
Hi
While custom calculations probably pose the most future-proof solution, as Kevin explains in his post, here are some alternatives you could use on the short term, depending on the use case:
- Conduct a Value Based Search on a TM_hour tag using the “constant” condition. Next, add a calculation for the average of your tag of interest. Using the scatter plot in Event Analytics, you can plot this average against the event start time and inspect the correlation. More information on time tags in this topic:
- Alternatively, you can use the import functionality in Tag Builder to generate a dummy tag that just increases with time over the period you are interested in. Load this tag together with you tag of interest and switch over to the scatter plot in your view to look into the regression equation.
Kind regards,
Jef
Hi Yves - a great question that deserves its own functional tips & tricks post. Please see here for a potential solution to your question!
Performing Linear Regression vs Time in TrendMiner
Kevin
Hi
While custom calculations probably pose the most future-proof solution, as Kevin explains in his post, here are some alternatives you could use on the short term, depending on the use case:
- Conduct a Value Based Search on a TM_hour tag using the “constant” condition. Next, add a calculation for the average of your tag of interest. Using the scatter plot in Event Analytics, you can plot this average against the event start time and inspect the correlation. More information on time tags in this topic:
- Alternatively, you can use the import functionality in Tag Builder to generate a dummy tag that just increases with time over the period you are interested in. Load this tag together with you tag of interest and switch over to the scatter plot in your view to look into the regression equation.
Kind regards,
Jef
+2Jef,
Thank you for providing these alternatives. I tried the Valuer-Based search against the TM_Hour of my time zone and this works very well. I will certainly also try the other trick as soon as I have a chance.
I very much appreciate all the support I get from everyone at Trendminer.
+2I tested the process suggested by Jef and created a new numeric tag using a .csv import that increments by 1 every hour from our system zero date with a projection into the future. Adding this tag to a TrendHub view allowed me to use a scatter plot and request any regression. This works very well!
Thank you
I’m glad you were able to implement this alternative approach Yves! Thank you Jef for a great suggestion!
+2Jef,
I clearly understand the options and we have implemented the one you suggested yourself i.e. importing a time tag using the .csv function with a projection into the future. We are satisfied that this workaround works in our context. Our users and management have advocated that this should be integrated as a built-in function of the product rather than a workaround, thus the enhancement request
Yves
Login to the community
No account yet? Create an account
Enter your E-mail address. We'll send you an e-mail with instructions to reset your password.