Little variation in latent variables

[wlangera]

Hi

I do not understand why the second latend variable explains almost no variation in my final model.
The ordination plot also looks strange and I have no idea what is going on.
There is only variation along LV1, but variance partioning shows only variation along LV2.

Data

See also #233

Response data:

> head(butterflies_data_wide[, -1])
# A tibble: 6 × 32
  `Bont zandoogje` Boomblauwtje `Bruin blauwtje` `Bruin zandoogje` Citroenvlinder Dagpauwoog `Gehakkelde aurelia`
             <dbl>        <dbl>            <dbl>             <dbl>          <dbl>      <dbl>                <dbl>
1                2            3                1                 5              6          5                    2
2                0            3                0                 1              5          0                    0
3                0            0                0                 2              1          0                    1
4                0            0                0                 0              1          1                    0
5                1            1                0                 0              3          0                    0
6                1            0                0                15              8          3                    0
# ℹ 25 more variables: `Groot dikkopje` <dbl>, `Grote vos` <dbl>, Hooibeestje <dbl>, Icarusblauwtje <dbl>,
#   `Klein koolwitje` <dbl>, `Kleine ijsvogelvlinder` <dbl>, `Kleine vos` <dbl>, `Kleine vuurvlinder` <dbl>,
#   Koninginnenpage <dbl>, Landkaartje <dbl>, `Oranje zandoogje` <dbl>, Oranjetipje <dbl>, Eikenpage <dbl>,
#   `Klein geaderd witje` <dbl>, Zwartsprietdikkopje <dbl>, Groentje <dbl>, `Groot koolwitje` <dbl>,
#   `Bont dikkopje` <dbl>, Heideblauwtje <dbl>, `Kleine parelmoervlinder` <dbl>, Heivlinder <dbl>,
#   Gentiaanblauwtje <dbl>, `Bruine eikenpage` <dbl>, Koevinkje <dbl>, Kommavlinder <dbl>

Site data:

> head(plot_data_wide)
# A tibble: 6 × 7
  habID habClass  habAreaM2 samplingEffort      x       y offset_var
  <fct> <fct>         <dbl>          <dbl>  <dbl>   <dbl>      <dbl>
1 6     dry heath    20394.            105 51116. 169014.       14.6
2 28    dry heath    13584.            101 78291. 196573.       14.1
3 29    dry heath    13642.             21 78373. 196689.       12.6
4 35    dry heath    15100.              7 74941. 198768.       11.6
5 39    dry heath    17703.             10 76287. 199974.       12.1
6 42    dry heath    18619.             47 75019. 200256.       13.7

Models

1. Only responses

# Response only
my_model1 <- gllvm(
  y = butterflies_data_wide[, -1],
  num.lv = 2,
  family = "ZINB",
  sd.errors = FALSE
)

# Convergence
hist(my_model1$TMBfn$gr())

# Create ordination
ordiplot(my_model1, symbols = TRUE, biplot = TRUE,
         s.colors = col_vect, s.cex = 0.5, pch = 16)
legend(
  "topright",
  legend = levels(plot_data_wide$habClass),
  col = my_colors,
  pch = 16,
  cex = 1
)

# Variance partitioning
plot_variance_partitioning(my_model1)

2. Categorical predictor and offset

# Categorical predictor and offset
my_model4 <- gllvm(
  y = butterflies_data_wide[, -1],
  X = plot_data_wide,
  formula = ~habClass,
  offset = plot_data_wide$offset_var,
  num.lv = 2,
  family = "ZINB",
  sd.errors = FALSE
)

# Convergence
hist(my_model2$TMBfn$gr())

# Create ordination
ordiplot(my_model4, symbols = TRUE, biplot = TRUE,
         s.colors = col_vect, s.cex = 0.5, pch = 16)
legend(
  "topright",
  legend = levels(plot_data_wide$habClass),
  col = my_colors,
  pch = 16,
  cex = 1
)

# Variance partitioning
plot_variance_partitioning(my_model4)

[BertvanderVeen]

Do you have your dataset somewhere available online so I may poke at it?

[wlangera]

I will see what I can do. I will get back to you next week

[wlangera]

I have send you a mail with a reproducible example and the data as csv. I cannot share it or put it online.