Dear Sujayanand G. K.
When data have a lot of zeros and is skewed in that direction. Statisticians have suggested using transformation using data point plus 0.5 square root. This will help normalize the data for analysis.
There are also tests to conduct to determine if your transformation were effective. Once effectively transformed in normalized distribution the data can be analyzed for variance and the means separated by appropriate statistics.
Fischers Least Significant Difference and Duncans New Multiple Range Test were used when I was conducting trials.
You may want to consult a competent stats man to get the latest.
Hope some of these ideas are useful to get started.
Dear Sujayanand G. K.
If data transformation do not work, the most accurate method for analysing count data is to use a generalized model (https://en.wikipedia.org/wiki/Generalized_linear_model). Such models do not rely on the normality assumption.
You can perform these analysis either with SAS Genmod procedure or with R glm (https://stat.ethz.ch/R-manual/R-devel/library/stats/html/glm.html).
Depending on wether over-dispersion is not present or present in your data (https://cran.r-project.org/web/packages/pscl/vignettes/countreg.pdf), you will have to use respectively either (1) a Poisson link function or (2) a negative binomial link function (with SAS) or a quasi-Poisson (with R).
first determine the percent pod damaged ( no.damaged/ total pods) . for stat. analysis add 0.5 to all values including zero values, transform the values and do the analysis to determine the LSD
- Badges
- Science topic
- Similar topics
- Invertebrate Zoology
- Entomology
- Insect
- Neoptera
- Holometabola
- Moths