Dear Roman

Do you mean you want to check the contribution of temperature and humidity on protein and fat in milk?

Am I right?

Dear Maryam

You correctly understood me. My goal is to assess how changes in the temperature and humidity index affect the yield of milk, fat and protein in cows during the hot summer period.

Thank you for your interest

There is a nice paper on this topic from the Journal of Dairy Science and I have attached a PDF of that paper.

Dear Jack

Many thanks for the response. I will certainly get acquainted with the material of this article

We compared weather predictions for cow heat stress (THI - a week ahead) and got good results to the measurements.

Dear Harel

I am grateful to you for your reply. The Israeli experience and achievements in dairy farming are impressive. I will follow your project

Construct a multiple linear regression model having a single dependant variable (milk/fat/protein) and independant variables-humidity/low/high temp to decide the relative importance of each and contribution too. Find out R-square for each prediction equation to decide the fitness of the model adopted. Decide if any transformation of the data is necessary to improve the R-square value.

Thanks

Dear Loknath G.R.

I will definitely follow your advice. I'm grateful for your response

I'd suggest contacting Ignacy Misztal at the University of Georgia. His group has done extensive research on genetics of heat tolerance (https://www.researchgate.net/project/Genetics-of-heat-tolerance).

Dear Suzanne

I am thankful for your answer. I have already read some of these scholars. I'm glad that you responded

That's a question which covers more than your specific topic and is of tremendous importance in all scientific topices. We always have to remember that in many research models are only a way to describe the observed data. Their fit is therefore much more optimist than when trying to use these parameters to predict future observation.

In general internal validation techniques (eg bootstrap and different cross-validation) are a key component of validation to use the estimates with more realistic prevision.The quality of prediction depends on the regression coefficient shrinkage that was obtained during internal validation techniques (model updating). I may suggest the textbook of Dr Ewout Steyerberg on that topic Clinical Prediction Model ( https://www.springer.com/gp/book/9780387772431 using R as a stat software).

Fitting models using a Bayesian framework is also a nice way to predict future data...

another aspect that needs to be covered is the "farm" or any other random effect that needs to be accounted for. In many cases, if our estimate is a "cow" specific estimate a specific farm effect (modifying the intercept (and eventually the slope of your coefficient)) that needs to be accounted for. The effect of the same parameters may vary depending on specific non measured herd characteristics.

Dear colleague. Thanks for the exhaustive answer, as well as a useful link. I will take this into account during research