issue with estimate_contrasts

[goesjon]

Hello,

I have the following problem with the function "estimate_contrasts()":

model <- lmer(data = Data, observationVariable ~ factor1 * factor2 + (1|subjectID))
{no errors or warnings here}
contrasts <- estimate_contrasts(model, levels = "factor1", fixed = "factor2")
{Error in names(levelcols) <- c("Level1", "Level2") :
'names' attribute [2] must be the same length as the vector [0]}

No matter what I do, I always get this error using "estimate_contrasts()" with the "fixed" option. What do I do wrong? Why does the vector have a length of [0]?

Help would be very much appreciated.

Thank you very much and stay healthy!

[DominiqueMakowski]

reproducible code:

df <- iris
df$factor1 <- ifelse(df$Sepal.Width > 3, "A", "B")
df$factor2 <- ifelse(df$Petal.Length > 3.5, "C", "D")

# Doesn't work
model <- lm(Petal.Width ~ factor1 * factor2, data=df)
estimate_contrasts(model, fixed = "factor2")

# Works ??
model <- lm(Petal.Width ~ Species * factor2, data=df)
estimate_contrasts(model, fixed = "factor2")

I will investigate

[DominiqueMakowski]

Could try updating from dev devtools::install_github("easystats/modelbased", "dev") and see if that works?

[goesjon]

Hello @DominiqueMakowski

thank you very much for your messages and for you reproducible code. Indeed, the "# Doesn't work" part does not work, whereas the "# Works ??" part works. I updated with:

devtools::install_github("easystats/modelbased", "dev")

but it still does not work.

I found a kind of work around:

estimate_contrasts(model, levels = ~ "factor1 * factor2")

but it is not really the same. Tell me, when you found out something.

Thank you very much for your help.

Jonathan

[DominiqueMakowski]

As the #doesn't work code now works at my place, could you reprex what's happening on your computer, including the error message?

library(modelbased)
df <- iris
df$factor1 <- ifelse(df$Sepal.Width > 3, "A", "B")
df$factor2 <- ifelse(df$Petal.Length > 3.5, "C", "D")

# Now works
model <- lm(Petal.Width ~ factor1 * factor2, data=df)
estimate_contrasts(model, fixed = "factor2")
#> Level1 | Level2 | factor2 | Difference |   SE |         95% CI |     t |  df |      p | Difference (std.)
#> ---------------------------------------------------------------------------------------------------------
#> A      |      B |       C |       0.40 | 0.08 | [ 0.25,  0.55] |  5.29 | 146 | < .001 |              0.52
#> A      |      B |       D |      -0.26 | 0.10 | [-0.46, -0.06] | -2.54 | 146 | 0.012  |             -0.34

^{Created on 2020-04-15 by the reprex package (v0.3.0)}

[goesjon]

Hello @DominiqueMakowski

thank you very much for your help. The "#doesn't work" part now (suddenly, I am really sorry) works also at my place. Also my original code (lmer model) now works. Great!

Thank you very much again. I close the issue.

[franrodalg]

Hi @DominiqueMakowski,

Sorry for resurfacing this, but I have just installed the development version because I was facing the same issue with my data, and I don't get the expected output for

estimate_contrasts(iris_fit, fixed = "factor2")

Instead, I get contrasts for a single level of the "fixed" factor:

Level1 | Level2 | factor2 | Difference |   SE |         95% CI |     t |  df |     p | Difference (std.)
--------------------------------------------------------------------------------------------------------
A      |      B |       D |      -0.26 | 0.10 | [-0.46, -0.06] | -2.54 | 146 | 0.012 |             -0.34

I tried with my own data and it causes the same behaviour. With multiple "fixed" factors, I only get a level from each.

Thanks so much for your work.

Cheers,

• Fran

[DominiqueMakowski]

@franrodalg could you share a reproducible example so I can investigate? thanks!

[franrodalg]

thanks for replying so quickly @DominiqueMakowski

I have run exactly the same code as above:

df <- iris
df$factor1 <- ifelse(df$Sepal.Width > 3, "A", "B")
df$factor2 <- ifelse(df$Petal.Length > 3.5, "C", "D")

# Doesn't work
model <- lm(Petal.Width ~ factor1 * factor2, data=df)
estimate_contrasts(model, fixed = "factor2")

[DominiqueMakowski]

What would you expect? Here it is rather expected, as essentially by default the contrast will be made between the factor1 condition (so between A and B), and factor2 is fixed, meaning maintained at one value. You can pass factor2 to modulate, in which case it will show the contrast at each value of factor2:

> estimate_contrasts(model, modulate = "factor2")
Level1 | Level2 | factor2 | Difference |   SE |         95% CI |     t |  df |      p | Difference (std.)
---------------------------------------------------------------------------------------------------------
A      |      B |       C |       0.40 | 0.08 | [ 0.25,  0.55] |  5.29 | 146 | < .001 |              0.52
A      |      B |       D |      -0.26 | 0.10 | [-0.46, -0.06] | -2.54 | 146 | 0.012  |             -0.34

Hope that clarifies

[franrodalg]

I was expecting the same output that you posted above:

As the #doesn't work code now works at my place, could you reprex what's happening on your computer, including the error message?

library(modelbased)
df <- iris
df$factor1 <- ifelse(df$Sepal.Width > 3, "A", "B")
df$factor2 <- ifelse(df$Petal.Length > 3.5, "C", "D")

# Now works
model <- lm(Petal.Width ~ factor1 * factor2, data=df)
estimate_contrasts(model, fixed = "factor2")
#> Level1 | Level2 | factor2 | Difference |   SE |         95% CI |     t |  df |      p | Difference (std.)
#> ---------------------------------------------------------------------------------------------------------
#> A      |      B |       C |       0.40 | 0.08 | [ 0.25,  0.55] |  5.29 | 146 | < .001 |              0.52
#> A      |      B |       D |      -0.26 | 0.10 | [-0.46, -0.06] | -2.54 | 146 | 0.012  |             -0.34

Created on 2020-04-15 by the reprex package (v0.3.0)