Here is the ideal (and currently unattainable) answer:

There are a number of problems to overcome, and these are typically ignored or glossed over. In part the data are very hard to get, and the analyses become more complicated than can be handled with an introductory course in statistics.

You need to figure out how the pesticide is distributed in the environment relevant to the life stage and size of the target/non-target insects. This distribution will interact with insect behavior to result in a toxicant acquisition rate in mg ai/hour.

You need to convert the dose mortality assay data into a mg ai/kg body weight.

Finally you need the rate at which the toxicant is metabolized.

Now you can figure out if the acquisition rate is greater than the degradation rate and how long it will take the insect to acquire a biologically relevant dose.

However, the insect changes over time, so these results will have to be linked to growth model. The dose will not affect all individuals equally, so you will need a population growth rate model.

To be complete, you might also add a model for crop growth rate and combine that with injury per insect and merge that with the insect (individual and population) growth models.

As far as I know, this has not been done for any insect in any crop. Bits and pieces have been done on various pests/crops. One of the biggest problems is to have analytical techniques appropriate for the small scale relevant to most insects. The distribution of the pesticide will be a function of the application technique and the way that it interacts with crop topology to influence the distribution of spray throughout the field.

Thanks for your detailed answer. The problem here is that I'm working with different taxa (and hundreds of species, from plants, carabid to birds...) in field plots over the whole country.... so it's going to be tricky if not impossible.

My main problem is to combine in an index different molecules. Unfortunately, and though this isn't really satisfying, I'll have to stick to the treatment frequency index...

Thanks again for taking the time to answer