1. This is a blog where you can post your doubts in mathematics and statistics and solutions to the queries will be posted immediately.
2. Sometimes I will post some good mathematical questions where you can utilize your brain to solve those problems which will help you in future.
3. I will also post some tricks in Mathematics.
4. If you know any tricks you can post it in the blog.
All the Best.......
Uff, finally I was able to find out how to post in a blog!
Dear Surya4u,
Im using R platform for statistical calculations, hope you know it. I have the following statistical problem:
I want to predict differences in potential to be captured between different lizard species due to their characteristics of speed, size and color.
For that, I generated a game which made people play and try to catch 50 lizards during each game session. The speed, size and color of those lizards and was saved to my computer, and also where the lizards were hit during the game(body=2, tail=1,fail=0). This is the variable result
So, I constructed a table with these data and calculated the combination of factors that best explained results across all the lizards played across all the game sessions. I did that fitting iteratively multinomial models to different combinations of the factors.
Once I reached the best combination (or the best model)
(for clarity, lets suppose that the best model is result= speed+body size:body color where"+" separates different terms and ":" means interaction among two terms )
Now, I want to use that model to predict the differences in catchability of my lizards now using real data from them (i.e. their speed, their body size and their body color)
I think that the most natural way to act now, would be to fit the best model with the game data, so as to get the coefficients for each term in the model, and then substitute with the real data.
However, this would give me values like 0, 1 or 2 for each species, as the model fitted is for a categorical ordered variable.
I would prefer to generate a continuous value, that allows me to find more subtle differences between the lizards (e.g. 0.8 versus 0.3)
Do you see a way to solve this problem? If you work with R, I could send you the code.
Uff, finally I was able to find out how to post in a blog!
ReplyDeleteDear Surya4u,
Im using R platform for statistical calculations, hope you know it. I have the following statistical problem:
I want to predict differences in potential to be captured between different lizard species due to their characteristics of speed, size and color.
For that, I generated a game which made people play and try to catch 50 lizards during each game session. The speed, size and color of those lizards and was saved to my computer, and also where the lizards were hit during the game(body=2, tail=1,fail=0). This is the variable result
So, I constructed a table with these data and calculated the combination of factors that best explained results across all the lizards played across all the game sessions. I did that fitting iteratively multinomial models to different combinations of the factors.
Once I reached the best combination (or the best model)
(for clarity, lets suppose that the best model is
result= speed+body size:body color
where"+" separates different terms and ":" means interaction among two terms
)
Now, I want to use that model to predict the differences in catchability of my lizards now using real data from them (i.e. their speed, their body size and their body color)
I think that the most natural way to act now, would be to fit the best model with the game data, so as to get the coefficients for each term in the model, and then substitute with the real data.
However, this would give me values like 0, 1 or 2 for each species, as the model fitted is for a categorical ordered variable.
I would prefer to generate a continuous value, that allows me to find more subtle differences between the lizards (e.g. 0.8 versus 0.3)
Do you see a way to solve this problem?
If you work with R, I could send you the code.
Many thanks
Agus