Linearize#
Create a linearize model based on a given input model.
The linearization procedure will produce and run two new models. Given an input model, a derivative model is created which will extract all derivatives for ETAs and EPSILONs. This is followed by the creation of a linearized model with input values of the ETAs updated according to the results from the derivative model.
Running#
To create a linearized model, please run
from pharmpy.modeling import load_example_model
from pharmpy.tools import run_linearize
start_model = load_example_model("pheno")
linres = run_linearize(start_model)
linearized_model = linres.final_model
start_model <- load_example_model("pheno")
linres <- run_linearize(start_model)
linearized_model <- linres$final_model
Arguments#
Argument |
Description |
|---|---|
|
Input model to linaerize |
|
Name to use for the linearized model. Default is “linbase” |
|
New description for linaerized model. Default is “” |
De-linearization#
A linearized model can be de-linearized as well. For this, a base model is required for parameters not stored in the linearized model. Such as thetas.
from pharmpy.tools import delinearize_model
start_model = read_model('path/to/model')
res = run_linearize(start_model)
linearized_model = res.final_model
delinearized_model = delinearize_model(linearized_model, start_model)
start_model <- read_model('path/to/model')
res <- run_linearize(start_model)
linearized_model <- res$final_model
delinearized_model <- delinearize_model(linearized_model, start_model)
For this tool, the option param_mapping can also be set, if the ETAs from the linearized
model should be mapped to some other parameter in the base model. An example of this can be
seen below.
Note
If param_mapping is used, then all ETAs in the linearized model are expected to
be mapped to a parameter.
from pharmpy.modeling import load_example_model
from pharmpy.tools import delinearize_model
start_model = load_example_model("pheno")
linres = run_linearize(start_model)
linearized_model = linres.final_model
param_mapping = {"ETA_1":"V", "ETA_2": "CL"}
delinearized_model = delinearize_model(linearized_model, start_model)
start_model <- load_example_model("pheno")
linres <- run_linearize(start_model)
linearized_model <- linres$final_model
param_mapping <- {"ETA_1":"V", "ETA_2": "CL"}
delinearized_model <- delinearize_model(linearized_model, start_model)
The linearize results#
OFVs#
The OFVs of the input model and the linearized model before and after estimation are summarized in the ofv table. These values should be close. A difference signals problems with the linearization.
| ofv | |
|---|---|
| base | 730.894727 |
| lin_evaluated | 730.894727 |
| lin_estimated | 730.847272 |
Individual OFVs#
The individual OFVs for the base and linearized models together with their difference is in the iofv table. If there was a deviation in the ofv these values can be used to see if some particular individual was problematic to linearize.
| base | linear | delta | |
|---|---|---|---|
| ID | |||
| 1 | 7.742853 | 7.722681 | -0.020172 |
| 2 | 12.049270 | 12.072922 | 0.023652 |
| 3 | 12.042005 | 12.025071 | -0.016933 |
| 4 | 12.812731 | 12.767326 | -0.045405 |
| 5 | 10.092668 | 10.052741 | -0.039927 |
| 6 | 14.345523 | 14.466217 | 0.120694 |
| 7 | 11.092993 | 11.062696 | -0.030297 |
| 8 | 13.515740 | 13.483016 | -0.032724 |
| 9 | 15.320532 | 15.253131 | -0.067401 |
| 10 | 10.998789 | 10.959488 | -0.039301 |
| 11 | 5.216717 | 5.214554 | -0.002163 |
| 12 | 12.099921 | 12.125228 | 0.025306 |
| 13 | 10.321679 | 10.306275 | -0.015405 |
| 14 | 18.261241 | 18.333707 | 0.072466 |
| 15 | 7.671243 | 7.651480 | -0.019763 |
| 16 | 12.330720 | 12.297562 | -0.033158 |
| 17 | 12.936161 | 12.906483 | -0.029678 |
| 18 | 19.714069 | 19.871090 | 0.157021 |
| 19 | 12.019825 | 12.011817 | -0.008008 |
| 20 | 12.056142 | 12.013498 | -0.042644 |
| 21 | 12.248747 | 12.213924 | -0.034823 |
| 22 | 7.605213 | 7.571494 | -0.033719 |
| 23 | 19.815937 | 19.898911 | 0.082974 |
| 24 | 27.454128 | 27.483765 | 0.029637 |
| 25 | 27.964631 | 28.119373 | 0.154742 |
| 26 | 13.186715 | 13.170103 | -0.016612 |
| 27 | 9.077661 | 9.064019 | -0.013642 |
| 28 | 7.940635 | 7.941894 | 0.001260 |
| 29 | 5.074883 | 5.073446 | -0.001437 |
| 30 | 9.256369 | 9.245489 | -0.010881 |
| 31 | 5.103887 | 5.101956 | -0.001932 |
| 32 | 19.907728 | 19.900485 | -0.007243 |
| 33 | 7.743720 | 7.709980 | -0.033740 |
| 34 | 8.047324 | 8.021009 | -0.026315 |
| 35 | 9.430306 | 9.400894 | -0.029412 |
| 36 | 13.781609 | 13.798011 | 0.016402 |
| 37 | 8.378940 | 8.371430 | -0.007511 |
| 38 | 16.194729 | 16.237151 | 0.042422 |
| 39 | 15.599213 | 15.525654 | -0.073559 |
| 40 | 6.709166 | 6.667521 | -0.041645 |
| 41 | 11.219054 | 11.180075 | -0.038978 |
| 42 | 18.122738 | 18.296506 | 0.173768 |
| 43 | 6.229686 | 6.228531 | -0.001156 |
| 44 | 10.756406 | 10.734272 | -0.022134 |
| 45 | 10.979740 | 10.927830 | -0.051910 |
| 46 | 4.813988 | 4.812185 | -0.001803 |
| 47 | 6.234962 | 6.233799 | -0.001163 |
| 48 | 35.389988 | 35.431669 | 0.041681 |
| 49 | 12.057167 | 12.047647 | -0.009521 |
| 50 | 19.429917 | 19.365265 | -0.064652 |
| 51 | 15.011212 | 15.105276 | 0.094064 |
| 52 | 16.302735 | 16.342891 | 0.040156 |
| 53 | 9.292516 | 9.307608 | 0.015092 |
| 54 | 15.067193 | 14.977095 | -0.090098 |
| 55 | 4.359971 | 4.357379 | -0.002592 |
| 56 | 7.340768 | 7.341140 | 0.000372 |
| 57 | 9.515376 | 9.511708 | -0.003668 |
| 58 | 11.970486 | 11.940640 | -0.029846 |
| 59 | 13.638461 | 13.592266 | -0.046195 |
This is also plotted in iofv_plot