In insecticide screening trials; sometimes live post count increases more than live pre-count resulting live count reduction value being a minus value. I entered live count reduction arcsign value to SAS for t test analysis with that minus value. The program interpret the results anyway.
1. Is it correct?
2. To avoid minus values a constant value can be added to all the data. Is it correct to do that?
Instead of using 'reduction' as your variable, you could just use absolute numbers before and after and then you won't get negative numbers. Or add the constant before doing a transformation :)
Logit transformation might work better, though you will have to deal with negative mortality before trying the analysis. I assume that some of the insects reproduced, and reproduction was greater than mortality. I might remove/discount new individuals added to a treatment. If all the original insects die, then mortality is 100% even if there are dozens of immatures left by the adults. I would note this fact as something for further study. Some pesticides are known to boost reproduction in other pests.
Just because the statistical package gives you numbers (as opposed to an error) does not mean that the numbers are correct. Bad data in = bad results out.
Might have to resort to alternative analysis approaches. Search for manuscripts on modeling dose-mortality for an herbicide called glyphosate. At low doses glyphosate promotes plant growth, at greater doses it kills plants, and at very high doses it burns the weed, but the weed recovers. This is not the typical dose-response curve that one expects.
- Badges
- Science topic
- Similar topics
- Correction