estimate change in R of linear model

Question

I'm doing the statistical analysis for my thesis with a linear mixed model, using the nlme package. Now I work with several orientations (west, east, north), and R takes the estimate as the alphabetically first value. However, I want the estimate to be the lowest value. Does somebody know how to do this?

Welcome to SO. Please read how to ask a good question **[here](https://stackoverflow.com/help/how-to-ask) ** — KoenV, May 24 '17 at 13:03
By "lowest value" do you mean the orientation in which the expected value of the response variable is lowest? — Ben Bolker, May 24 '17 at 13:20
I have an idea for how to do what I *think* you want, but you should clarify your question first please ... — Ben Bolker, May 24 '17 at 14:26

score 1 · Answer 1 · answered May 24 '17 at 17:35

Like @BenBolker, I think I know what you have in mind.

First, simulating some toy data...

site <- rep(1:3, each=6)
orientation <- rep(c("west","east","north"), 6)
x <- rnorm(18)
x[orientation=="north"] <- x[orientation=="north"] + 1
x[orientation=="east"] <- x[orientation=="east"] + 2
df <- data.frame(site, orientation, x)
df
#    site orientation          x
# 1     1        west  0.9554461
# 2     1        east  4.1742621
# 3     1       north  0.8901404
# 4     1        west -0.5998719
# 5     1        east  1.8279064
# 6     1       north  1.0984336
# 7     2        west  1.1020786
# 8     2        east  2.7994980
# 9     2       north  0.9158762
# 10    2        west  1.2512472
# 11    2        east  0.1200988
# 12    2       north  1.3584406
# 13    3        west  0.6602157
# 14    3        east  0.6226468
# 15    3       north  0.9409308
# 16    3        west -1.8992221
# 17    3        east  2.0386123
# 18    3       north  2.3562818

Running lme() yields...

library(nlme)
summary(lme(x~orientation, random=~1|site, data=df))
# .... just looking at fixed effects ...
# Fixed effects: x ~ factor(orientation) 
#                       Value Std.Error DF   t-value p-value
# (Intercept)       1.9305041 0.4729373 13  4.081945  0.0013
# orientationnorth -0.6704868 0.6688344 13 -1.002471  0.3344
# orientationwest  -1.6855218 0.6688344 13 -2.520088  0.0256

... and you'd like west to be the reference level instead of east, since it's got the lowest effect.

One possible solution: use a factor version of your orientation variable, with the levels in the order you'd like them to be in.

df$orientation1 <- factor(df$orientation, levels=c("west","north","east"))
summary(lme(x~orientation1, random=~1|site, data=df))
# ... again just looking at fixed effects ...
#                       Value Std.Error DF   t-value p-value
# (Intercept)       0.2449823 0.4729373 13 0.5180016  0.6132
# orientation1north 1.0150350 0.6688344 13 1.5176178  0.1530
# orientation1east  1.6855218 0.6688344 13 2.5200884  0.0256

estimate change in R of linear model

1 Answers1