1
$\begingroup$

The summary of my GLM shows day 12 of factor Date to be significant, but anova(model, test="Chisq") shows Date to be not significant overall. I know how to report the statistics from the Chisq table, but as I have z values in the summary table I am unsure how to report, or if I should report, that day 12 is significant.

Similarly when finding that Date is significant, how to report what specific dates seem to be important.

Thanks in advance

Lara

I have Fertility and Fecundity of female flies measured over 12 days. I want to check if there is a decline (or otherwise) in fertility/fecundity over this time.

Additionally, at day 13 females are mated (with a male from one of two different groups), and fertility and fecundity are measured until day 20. I want to use Date as a factor to identify significant peaks in fertility/fecundity i.e. after mating, and potential difference in peaks between groups.

Call:
glm(formula = dda$Fertility.Absolute ~ sqrt(dda$Fecundity) + 
    dda$Group + dda$Date + sqrt(dda$Fecundity):dda$Group + sqrt(dda$Fecundity):dda$Date, 
    family = poisson)

Deviance Residuals: 
    Min       1Q   Median       3Q      Max  
-3.1397  -0.6786  -0.4797   0.3596   3.7588  

Coefficients:
                                   Estimate Std. Error z value Pr(>|z|)    
    (Intercept)                    -1.91501    0.51539  -3.716 0.000203 ***
    sqrt(dda$Fecundity)             0.72372    0.12441   5.817 5.99e-09 ***
    dda$Group2                      0.19540    0.19230   1.016 0.309585    
    dda$Date4                       0.18117    0.62648   0.289 0.772439    
    dda$Date6                      -0.28952    0.68983  -0.420 0.674706    
    dda$Date8                       0.07111    0.60531   0.117 0.906480    
    dda$Date10                      0.19557    0.62232   0.314 0.753325    
    dda$Date12                      0.79619    0.60710   1.311 0.189696    
    dda$Date14                      1.93702    0.53938   3.591 0.000329 ***
    dda$Date16                      0.75623    0.58296   1.297 0.194554    
    dda$Date18                     -0.05392    0.67805  -0.080 0.936618    
    dda$Date20                     -0.26291    0.68841  -0.382 0.702530    
    sqrt(dda$Fecundity):dda$Group2 -0.07309    0.04822  -1.516 0.129583    
    sqrt(dda$Fecundity):dda$Date4   0.27388    0.17555   1.560 0.118734    
    sqrt(dda$Fecundity):dda$Date6   0.37684    0.22832   1.651 0.098836 .  
    sqrt(dda$Fecundity):dda$Date8   0.13017    0.13861   0.939 0.347674    
    sqrt(dda$Fecundity):dda$Date10  0.04552    0.15345   0.297 0.766722    
    sqrt(dda$Fecundity):dda$Date12 -0.16593    0.14861  -1.117 0.264186    
    sqrt(dda$Fecundity):dda$Date14 -0.24864    0.12754  -1.949 0.051240 .  
    sqrt(dda$Fecundity):dda$Date16  0.05496    0.14578   0.377 0.706170    
    sqrt(dda$Fecundity):dda$Date18  0.15439    0.19341   0.798 0.424715    
    sqrt(dda$Fecundity):dda$Date20 -0.02006    0.16314  -0.123 0.902161    
---
Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1 

(Dispersion parameter for poisson family taken to be 1)

    Null deviance: 1327.36  on 359  degrees of freedom
Residual deviance:  298.26  on 338  degrees of freedom
AIC: 873.65

Number of Fisher Scoring iterations: 5

###

Analysis of Deviance Table    
Model: poisson, link: log    
Response: dda$Fertility.Absolute    
Terms added sequentially (first to last)    

                                  Df Deviance Resid. Df Resid. Dev  Pr(>Chi)    
NULL                                            359    1327.36              
    sqrt(dda$Fecundity)            1   893.88       358     433.48 < 2.2e-16 ***
    dda$Group                      1     0.03       357     433.45    0.8699    
    dda$Date                       9    82.16       348     351.29  6.01e-14 ***
    sqrt(dda$Fecundity):dda$Group  1     0.07       347     351.22    0.7859    
    sqrt(dda$Fecundity):dda$Date   9    52.96       338     298.26  2.97e-08 ***
$\endgroup$
2
  • 1
    $\begingroup$ The first half of this is missing, i.e. what your scientific problem is, how Date features in your analysis, why you are treating it as a factor, etc. $\endgroup$
    – Nick Cox
    Commented Oct 15, 2013 at 16:42
  • $\begingroup$ Probably related $\endgroup$
    – Affine
    Commented Oct 15, 2013 at 17:05

0