We illustrate how to use JMP with the insulating fluid data to be imported below. These data gives the time in minutes to dielectric breakdown or removal of 76 batches of insulating fluid tested at the following voltages (kilovolts):
26KV, 28KV, 30KV, 32KV, 34KV, 36KV, 38KV
Import into JMP the insulating fluid data:
(Instructions for importing a text file into JMP)
Check the following JMP handout "Accelerated Life Model: Distribution Identification" to see how we concluded that a lognormal regression is appropriate and to see why we decided to exclude the data corresponding to 32KV. Check the handout "Accelerated Life Model: Lognormal Regression" to see how to this lognormal regression can be done using the much simpler Survival JMP platform. The advantage of the approach in this handout is that we get the variance-covariance matrix of estimated parameters.
The JMP table should look as follows:
The code 0 means a failure time, the code 1 a removal time. The Frequency columns gives the number of batches that were removed at a given time. Notice that the data corresponding to 32KV has been excluded. We have defined three columns using JMP's calculator.
Since voltage is the accelerating stress the inverse power law applies. Hence we define an new column "LogKV" with log voltage. We use JMP's calculator to define this column:
We define the column "mu" using JMP's calculator as follows:
We define the column "loss" using JMP's calculator as follows:
We proceed to use JMP's Nolinear Fit platform as follows:
On the resulting panel we check "Loss is -LogLikelihood." Then click on Reset and Go and then on Confid Limits we get the following output:
Compare with the results in "Accelerated Life Model: Lognormal Regression."