Hi, For my master thesis I have 24 micro-plots on which I did measurements during 3 months. The measurements were: - Rainfall and runoff events throughout 3monts (runoff being dependant on the rainfall, a coefficient (%) has been made per rainfall event and per 3 months) - Soil texture (3 different textures were differentiated) - Slope (3 classes of slopes) - Stoniness (one time measurement) - Random roughness (throughout 3 months) - Land use (crop land or grazing land) - Vegetation cover (throughout 3 months) - Vegetation height (throughout 3 months, only measured on cropland) - Antecedent moisture content (throughout 3 months) Now I would like to investigate the effect of all these variables on the rainfall/runoff. For example does a steeper slope have a larger effect on the runoff than the soil texture? What are the possibilities in R? Thank you for any feedback, Francesca [[alternative HTML version deleted]]
Francesca Sorbie <fsorbie <at> hotmail.com> writes:> > Hi, > > For my master thesis I have 24 micro-plots on which I > did measurements during 3 months. > > The measurements were:> - Rainfall and runoff events throughout 3monts> (runoff being dependant on the rainfall, a coefficient (%) > has been made per rainfall event and per 3 months) > - Soil texture (3 different textures were differentiated) > - Slope (3 classes of slopes) > - Stoniness (one time measurement) > - Random roughness (throughout 3 months) > - Land use (crop land or grazing land) > - Vegetation cover (throughout 3 months) > - Vegetation height (throughout 3 months, only measured on cropland) > - Antecedent moisture content (throughout 3 months)> Now I would like to investigate the effect of all these variables on > the rainfall/runoff. For example does a steeper slope have a larger > effect on the runoff than the soil texture? What are the > possibilities in R?I'm afraid this is an R list, not a statistical consulting list. You probably want to do multivariate linear regression with lm(), after appropriate graphical exploration of the data to make sure there aren't glaring nonlinearities. You probably need to cut down or condense your list of predictors somehow; the rule of thumb is that with n observations you should generally not try to fit a model with more than n/10 parameters. Don't be tempted by stepwise regression. Read Frank Harrell's book on Regression Modeling Strategies, and (!!!) get some local statistical help -- if you're a student, someone should be teaching you ... Ben Bolker
Try the pairs() function to explore your raw data. This webpage may even give you a nice way to visualize your data: http://r-epid.blogspot.com/2008/11/correlation-pairs-plot.html -Steve -----Original Message----- From: Francesca Sorbie [mailto:fsorbie at hotmail.com] Sent: Saturday, April 14, 2012 6:44 AM To: r-help at r-project.org Subject: [R] master thesis Hi, For my master thesis I have 24 micro-plots on which I did measurements during 3 months. The measurements were: - Rainfall and runoff events throughout 3monts (runoff being dependant on the rainfall, a coefficient (%) has been made per rainfall event and per 3 months) - Soil texture (3 different textures were differentiated) - Slope (3 classes of slopes) - Stoniness (one time measurement) - Random roughness (throughout 3 months) - Land use (crop land or grazing land) - Vegetation cover (throughout 3 months) - Vegetation height (throughout 3 months, only measured on cropland) - Antecedent moisture content (throughout 3 months) Now I would like to investigate the effect of all these variables on the rainfall/runoff. For example does a steeper slope have a larger effect on the runoff than the soil texture? What are the possibilities in R? Thank you for any feedback, Francesca [[alternative HTML version deleted]]