Tukey for GLM can't find data in model

Question

I fit a GLM to a dataset. Now I want to see where the difference between my groups is, so I tried to run a Tukey HSD as a post-hoc test.

Because of it is a GLM, I can't use TukeyHSD(). So I tried to run a Tukey test with:

summary(glht(my.mod, mcp(treatment="Tukey")))  

When I tried this, I get the following error:

Variable(s) ‘treatment’ have been specified in ‘linfct’ but cannot be found in ‘model’!

This is what it looks like in R:

> recol
   Sample At..Ind. At.Sp.   X treatment
1      C1        1      1   0         C
2      C2        2      2   0         C
3      C3        1      1   0         C
4     FS1        1      1  50        FS
5     FS2        4      3  80        FS
6     FS3        0      0  80        FS
7     FL1        3      3 280        FL
8     FL2        4      3 310        FL
9     FL3        2      2 220        FL
10    VS1        2      2  40        VS
11    VS2        0      0  50        VS
12    VS3        1      1   5        VS
13    VL1        3      3 400        VL
14    VL2        7      6 600        VL
15    VL3        3      3 500        VL
16    MS1        1      1  30        MS
17    MS2        3      3  30        MS
18    MS3        0      0  30        MS
19    ML1        2      1 400        ML
20    ML2        7      7 300        ML
21    ML3        7      7 300        ML
>  my.mod=glm(Arachnida~as.factor(treatment),data=soort)
> summary(my.mod)

Call:
glm(formula = Arachnida ~ as.factor(treatment), data = soort)

Deviance Residuals: 
    Min       1Q   Median       3Q      Max  
-2.3333  -0.3333   0.0000   0.0000   2.6667  

Coefficients:
                         Estimate Std. Error t value Pr(>|t|)   
(Intercept)             3.107e-16  6.901e-01   0.000  1.00000   
as.factor(treatment)FL -1.277e-16  9.759e-01   0.000  1.00000   
as.factor(treatment)FS -3.511e-16  9.759e-01   0.000  1.00000   
as.factor(treatment)ML  3.000e+00  9.759e-01   3.074  0.00825 **
as.factor(treatment)MS  3.333e-01  9.759e-01   0.342  0.73775   
as.factor(treatment)VL  2.333e+00  9.759e-01   2.391  0.03141 * 
as.factor(treatment)VS  3.333e-01  9.759e-01   0.342  0.73775   
---
Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

(Dispersion parameter for gaussian family taken to be 1.428571)

    Null deviance: 48.571  on 20  degrees of freedom
Residual deviance: 20.000  on 14  degrees of freedom
AIC: 74.571

Number of Fisher Scoring iterations: 2

> glht(my.mod, mcp(treatment="Tukey"))
Error in mcp2matrix(model, linfct = linfct) : 
Variable(s) ‘treatment’ have been specified in ‘linfct’ but cannot be found in ‘model’ 
> summary(glht(my.mod, mcp(treatment="Tukey")))
Error in mcp2matrix(model, linfct = linfct) : 
Variable(s) ‘treatment’ have been specified in ‘linfct’ but cannot be found in ‘model’ 

Could anyone help me with my problem?

Answer 1

I think because of as.factor.

I experienced same error message when using summary(glht(mod1, mcp(cycle="Tukey"))) for model:

mod1 <- glm(dt$presence ~ dt$cycle + dt$water)

After I changed the formula to be:

mod1 <- glm(presence ~ cycle + water, data=dt)

It works.

For your case, try to use the following formula:

my.mod=glm(Arachnida ~ treatment,data=soort)

Since treatment already use character, no need as.factor()