0
$\begingroup$

I have run an experiment investigating how fructose concentration changes for mosquitoes held in four different preservative methods over 3 time periods (days 7, 14 and 21).

Here's what my data looks like:

'data.frame':   120 obs. of  4 variables:
 $ treatment             : Factor w/ 5 levels "C_ETOH","Frozen",..: 5 5 5 5 5 5 5 5 1 1 ...
 $ day                   : int  7 7 7 7 7 7 7 7 7 7 ...
 $ mean_absorbance       : num  0.472 0.652 0.284 0.421 0.693 ...
 $ fructose_concentration: num  0.559 0.802 0.305 0.49 0.857 ...

The fructose concentration residuals did not violate linearity, so I fit a glm() to my data

PRES_data_glm <- glm(fructose_concentration ~ day*treatment, data = PRES_data)

Running an ANOVA on the model showed there was a significant effect of treatment and an interaction with day and treatment.

Anova(PRES_data_glm)
Analysis of Deviance Table (Type II tests)

Response: fructose_concentration
              LR Chisq Df Pr(>Chisq)    
day              3.415  1     0.0646 .  
treatment       65.576  4  1.946e-13 ***
day:treatment   41.057  4  2.616e-08 ***
---
Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

I want to use emmeans to examine the effect of each preservation method on each day and across days.

Unfortunately, many of the emmeans comparisons (and contrasts) I try only produce the day 14 comparison with the treatments.

comparisons_PRES <- emmeans(PRES_data_glm, pairwise ~ day*treatment)
comparisons_PRES

$`emmeans`
 day treatment emmean     SE  df asymp.LCL asymp.UCL
  14 C_ETOH     0.522 0.0451 Inf     0.434     0.611
  14 Frozen     0.845 0.0451 Inf     0.757     0.934
  14 HF         0.841 0.0451 Inf     0.752     0.929
  14 KOD        0.990 0.0451 Inf     0.901     1.078
  14 W_ETOH     0.654 0.0451 Inf     0.566     0.743

Confidence level used: 0.95 

$contrasts
 contrast              estimate     SE  df z.ratio p.value
 14,C_ETOH - 14,Frozen -0.32305 0.0638 Inf -5.061  <.0001 
 14,C_ETOH - 14,HF     -0.31834 0.0638 Inf -4.987  <.0001 
 14,C_ETOH - 14,KOD    -0.46721 0.0638 Inf -7.320  <.0001 
 14,C_ETOH - 14,W_ETOH -0.13202 0.0638 Inf -2.068  0.2339 
 14,Frozen - 14,HF      0.00471 0.0638 Inf  0.074  1.0000 
 14,Frozen - 14,KOD    -0.14415 0.0638 Inf -2.258  0.1586 
 14,Frozen - 14,W_ETOH  0.19103 0.0638 Inf  2.993  0.0232 
 14,HF - 14,KOD        -0.14886 0.0638 Inf -2.332  0.1347 
 14,HF - 14,W_ETOH      0.18632 0.0638 Inf  2.919  0.0289 
 14,KOD - 14,W_ETOH     0.33518 0.0638 Inf  5.251  <.0001 

P value adjustment: tukey method for comparing a family of 5 estimates 

I have tried the suggestions outlined in the answer of a previous post, Pairwise comparisons via emmeans

pairs(comparisons_PRES, simple = "each")

But it is only giving me the day 14 comparisons (I suspect it may be to do with how I have written the emmeans comparison). Any help would be much appreciate.

$\endgroup$
2
  • $\begingroup$ Please note in the model summary that day has only 1 d.f. That's because you have put it in the model as a numeric predictor, and apparently its average value is 14. Replace day by factor(day) in the data, re-fit the model, and re-do all the summaries and EMMs. $\endgroup$
    – Russ Lenth
    Jun 3, 2021 at 1:49
  • $\begingroup$ Thanks kindly @RussLenth. That has indeed solved the problem. $\endgroup$
    – mossie_tom
    Jun 3, 2021 at 5:20

0

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.