Differences

This shows you the differences between two versions of the page.

--- resources:lme [2015/06/24 11:18] – anthony
+++ resources:lme [2019/05/22 16:08] (current) – external edit 127.0.0.1
@@ Line 26: / Line 26: @@
 http://cran.r-project.org/doc/manuals/r-devel/R-data.html
-''Column names can be given explicitly via the col.names; explicit names override the header line (if present).''
+>Column names can be given explicitly via the col.names; explicit names override the header line (if present).
@@ Line 32: / Line 32: @@
 http://cran.r-project.org/web/packages/arm/index.html
-''arm: Data Analysis Using Regression and Multilevel/Hierarchical Models
-R functions for processing lm, glm, svy.glm, merMod and polr outputs.''
+>''arm: Data Analysis Using Regression and Multilevel/Hierarchical Models''
+>
+>R functions for processing lm, glm, svy.glm, merMod and polr outputs.
-**arm** package provides the coefplot() function for plotting the regression coefficient values and std. errors.
+**arm** package provides the ''coefplot'' function for plotting the regression coefficient values and std. errors.
-Need to specify that subjects is a nominal variable, using the //factor// command:
+Need to specify that subjects is a nominal variable, using the ''factor'' command:
 http://www.statmethods.net/input/datatypes.html
-''Tell R that a variable is nominal by making it a factor. The factor stores the nominal values as a vector of integers in the range [ 1... k ] (where k is the number of unique values in the nominal variable), and an internal vector of character strings (the original values) mapped to these integers. ''
+>Tell R that a variable is nominal by making it a factor. The factor stores the nominal values as a vector of integers in the range [ 1... k ] (where k is the number of unique values in the nominal variable), and an internal vector of character strings (the original values) mapped to these integers.
 ===== Interpretation of glmer output =====
@@ Line 50: / Line 52: @@
 https://stat.ethz.ch/pipermail/r-sig-mixed-models/2011q1/015590.html
+>In this case the coefficient corresponds to one of the
-In this case the coefficient corresponds to one of the
+>terms in the model and I would advocate performing a likelihood ratio
-terms in the model and I would advocate performing a likelihood ratio
+>test comparing the two models
-test comparing the two models
+>
+><code rsplus>
-<code rsplus>
+>fm <- glmer(SameSite~BreedSuc1+Sex+(1|Bird), family="binomial")
-fm <- glmer(SameSite~BreedSuc1+Sex+(1|Bird), family="binomial")
+>fm0 <- glmer(SameSite~Sex+(1|Bird), family="binomial")  # the null
-fm0 <- glmer(SameSite~Sex+(1|Bird), family="binomial")  # the null
+>hypothesis model
-hypothesis model
+>anova(fm0, fm)
-anova(fm0, fm)
+></code>
-</code>
+>Even though the function is called anova it will, in this case,
+>perform a likelihood ratio test (LRT).  It also prints the values of
-Even though the function is called anova it will, in this case,
+>AIC and BIC if you prefer to compare models according to one of those
-perform a likelihood ratio test (LRT).  It also prints the values of
+>criteria but I prefer using the likelihood ratio for nested models.
-AIC and BIC if you prefer to compare models according to one of those
+>However, before doing that comparison you should ask yourself whether
-criteria but I prefer using the likelihood ratio for nested models.
+>you want to compare models that have the, apparently unnecessary term
+>for Sex in them.  The way I would approach the model building is first
-However, before doing that comparison you should ask yourself whether
+>to reduce the model to
-you want to compare models that have the, apparently unnecessary term
+><code rsplus>
-for Sex in them.  The way I would approach the model building is first
+>fm1 <- lmer(SameSite~BreedSuc1+(1|Bird), family="binomial")
-to reduce the model to
+></code>
+>You could then compare
-<code rsplus>
+><code rsplus>
-fm1 <- lmer(SameSite~BreedSuc1+(1|Bird), family="binomial")
+>anova(fm1, fm)
-</code>
+></code>
+>which I presume will give a large p-value for the LRT, so we prefer
-You could then compare
+>the simpler model, fm1.  After that, I would compare
+><code rsplus>
-<code rsplus>
+>fm2 <- lmer(SameSite ~ 1 + (1|Bird), family="binomial")
-anova(fm1, fm)
+></code>
-</code>
-which I presume will give a large p-value for the LRT, so we prefer
-the simpler model, fm1.  After that, I would compare
-<code rsplus>
-fm2 <- lmer(SameSite ~ 1 + (1|Bird), family="binomial")
-</code>