How to correct shift data series

Question

Hi All,

Here i have some data from a kind of measurement device. Now due to transmission cause, occaionally we may run into shift(drift) of data series. The drifted data is about 2.0 lower than normal values. How i can make them back normal value. I attached the data table and with graph script. Previouslt i  used dif function to fix. But it seems not work since the shift influece not just one data. It influence data series.

GoodMan

“2018-02-07”的子集.jmp

GoodMan · Answer

Hi Jim, thanks for your help. I understand your solution. Also it works for certain point. But my problem as showed in photo involved many points. I do not know how to handle with distribution. The arrow points are points i want to bring back to main line.I attached the whole process.Thank.

GM

2018-02-07.jmp

txnelson · Answer

I assume, that what you are looking for, is 1. A way to detect if there is a drift 2. Which data points have drifted 3. How much to change those values to bring them back into alignment I believe the Distribution Platform can answer these questions for you. If you run a Distribution on Temp, and then select Continuous Fit==>Normal Mixtures==>Normal 2 Misture 1. The 2 distributions can clearly be seen 2. You can select the rows from the display to identify which ones need to be adjusted 3. If you look into the Parameter estimates for the Normal 2 Mixture, you can see the Mu for each of the distributions, and by subtracting the lower from the higher -30.80698-(-32.75722) = 1.95024 And here is a script that will change the selected values by adding 1.95024 to each value of Temp Names Default To Here( 1 ); dt = Current Data Table(); rowMat = dt << get selected rows; For( i = 1, i <= N Rows( rowMat ), i++, dt:temp[rowMat[i]] = dt:temp[rowMat[i]] + 1.95024 );