GLM and POST HOC test INTERPRETATION
Your questions are basically statistical and therefore OT here, although some kind soul may respond. I would strongly suggest that you consult with a local statistical expert, as you seem to be out of your depth statistically. Cheers, Bert Bert Gunter "The trouble with having an open mind is that people keep coming along and sticking things into it." -- Opus (aka Berkeley Breathed in his "Bloom County" comic strip ) On Wed, Feb 8, 2017 at 4:08 PM, CHIRIBOGA Xavier
<xavier.chiriboga at unine.ch> wrote:
Dear colleagues, I am analyzing a data set of 68 values (integers). In some treatments (exactly 6) the values are "zero". Because I record 0 in my measurement (or really a small value below zero) My experiment is designed in such a way that I record values for 6 treatments at 2 times. Replicates are different in each combination time-treatment. I am running a GLM , poisson distribution, for ANOVA I used Chisq, and for the POST HOC test I used Tukey. I try to detect if interaction is significant, so I build the script: expresion~time*treatment Effects of time, treatment are interaction are significant. However, when I run the script for Tukey comparisons, I only get 15 comparisons. Of course I cannot interpret that: these comparisons are the same for Time 1 and Time 2, since there is a significant effect of time. Moreover, I got a warning message : covariate interactions found. I dont know if I am doing right? I dont know what to do? Thank you for your help, Xavier PhD Student University of Neuchatel lm3=glm(expresion~time*treatment,family="poisson")
summary(lm3)
Call:
glm(formula = expresion ~ time * treatment, family = "poisson")
Deviance Residuals:
Min 1Q Median 3Q Max
-5.3796 -1.4523 -0.6642 1.2277 6.3909
Coefficients:
Estimate Std. Error z value Pr(>|z|)
(Intercept) 2.09964 0.29508 7.115 1.12e-12 ***
time 0.20294 0.19255 1.054 0.291895
treatmentCHA0+Db -0.17004 0.36180 -0.470 0.638356
treatmentDb 1.68952 0.37624 4.490 7.11e-06 ***
treatmentHEALTHY 0.84035 0.50340 1.669 0.095049 .
treatmentPCL 0.32072 0.37950 0.845 0.398041
treatmentPCL+Db 0.54365 0.34047 1.597 0.110320
time:treatmentCHA0+Db 0.87314 0.22626 3.859 0.000114 ***
time:treatmentDb -0.82803 0.26539 -3.120 0.001808 **
time:treatmentHEALTHY -1.36987 0.38318 -3.575 0.000350 ***
time:treatmentPCL 0.08474 0.24635 0.344 0.730851
time:treatmentPCL+Db 0.39244 0.21521 1.824 0.068217 .
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
(Dispersion parameter for poisson family taken to be 1)
Null deviance: 1173.05 on 66 degrees of freedom
Residual deviance: 403.07 on 55 degrees of freedom
AIC: 707.95
Number of Fisher Scoring iterations: 5
anova(lm3,test="Chisq")
Analysis of Deviance Table
Model: poisson, link: log
Response: expresion
Terms added sequentially (first to last)
Df Deviance Resid. Df Resid. Dev Pr(>Chi)
NULL 66 1173.05
time 1 100.55 65 1072.50 < 2.2e-16 ***
treatment 5 561.69 60 510.81 < 2.2e-16 ***
time:treatment 5 107.75 55 403.07 < 2.2e-16 ***
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
summary(glht(lm3, mcp(treatment="Tukey")))
Simultaneous Tests for General Linear Hypotheses
Multiple Comparisons of Means: Tukey Contrasts
Fit: glm(formula = expresion ~ time * treatment, family = "poisson")
Linear Hypotheses:
Estimate Std. Error z value Pr(>|z|)
CHA0+Db - CHA0 == 0 -0.1700 0.3618 -0.470 0.9970
Db - CHA0 == 0 1.6895 0.3762 4.490 <0.001 ***
HEALTHY - CHA0 == 0 0.8404 0.5034 1.669 0.5402
PCL - CHA0 == 0 0.3207 0.3795 0.845 0.9568
PCL+Db - CHA0 == 0 0.5437 0.3405 1.597 0.5892
Db - CHA0+Db == 0 1.8596 0.3135 5.931 <0.001 ***
HEALTHY - CHA0+Db == 0 1.0104 0.4584 2.204 0.2266
PCL - CHA0+Db == 0 0.4908 0.3174 1.546 0.6231
PCL+Db - CHA0+Db == 0 0.7137 0.2696 2.648 0.0817 .
HEALTHY - Db == 0 -0.8492 0.4699 -1.807 0.4491
PCL - Db == 0 -1.3688 0.3338 -4.101 <0.001 ***
PCL+Db - Db == 0 -1.1459 0.2887 -3.969 <0.001 ***
PCL - HEALTHY == 0 -0.5196 0.4725 -1.100 0.8764
PCL+Db - HEALTHY == 0 -0.2967 0.4418 -0.672 0.9842
PCL+Db - PCL == 0 0.2229 0.2929 0.761 0.9725
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
(Adjusted p values reported -- single-step method)
Warning message:
In mcp2matrix(model, linfct = linfct) :
covariate interactions found -- default contrast might be inappropriate
[[alternative HTML version deleted]]
______________________________________________ R-help at r-project.org mailing list -- To UNSUBSCRIBE and more, see https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.