MethodologyJakubHorák
Methodology
MethodologyJakubHorák
Methodology
• Datacollection.• Dataevaluation.
MethodologyJakubHorák
Datacollection
MethodologyJakubHorák
Datacollection
(1) Dataalready collected.• Datacollected inagood way (Yes!).• Datawrongly collected (Need tolearn statistics…).
(2) Datastill notcollected.• Doyou welcome an advice?(Yes!).• You areunreliable (Need tolearn stats…)
MethodologyJakubHorák
Datacollection
• There is noideal sampling design,however…– Regular grid.
MethodologyJakubHorák
Datacollection
• There is noideal sampling design,however …– Random.
MethodologyJakubHorák
Datacollection
• There is noideal sampling design,however …– Equal-stratified.
MethodologyJakubHorák
Datacollection
• There is noideal sampling design,however …– Proportional-stratified.
MethodologyJakubHorák
Datacollection
• There is noideal sampling design,however …
MethodologyJakubHorák
Datacollection
• (Pseudo)replication is notaproblem,butreviewers do…
MethodologyJakubHorák
Datacollection
• (Pseudo)replication is notaproblem,butreviewers do…
MethodologyJakubHorák
Datacollection
• (Pseudo)replication is justaproblem of space,which is aproblem that can be solved…(…sometimes quite hardly).
MethodologyJakubHorák
Dataevaluation
MethodologyJakubHorák
www.zahrady-iberis.cz
Dataevaluation
Descriptive statistics–Mean.–Median.–Modus.–Minimum.–Maximum.– Standarderror.– Standarddeviation…
MethodologyJakubHorák
Dataevaluation
(1)Testof dependent variable for spatialautocorrelation.– E.g.MoranIor Geary C…– P≥ 0.05 traditional stats methods can be used.– Software:SAM,R...
MethodologyJakubHorák
Dataevaluation
(1)Testof dependent variable for spatialautocorrelation.– P< 0.05 (significant)possibility touse:• Autocovariate (R).• SAR(R,SAM).• PCNM(R)…
MethodologyJakubHorák
Dataevaluation
(2)Testof dependent variable for normalityindistribution.– E.g.Combination of histogramandShapiro-Wilktest.
– P≥ 0.05 parametricstatsmethodscanbeused.– Software:Statistica,R.
MethodologyJakubHorák
Dataevaluation
(2)Testof dependent variable for normalityindistribution.– P< 0.05:• Datatransformation.• Non-parametric methods.• Testonother distributions (Poisson,Gamma,quasi-Poisson,Zero-inflated…).• Testof modelresiduals onnormality.
MethodologyJakubHorák
Dataevaluation
(3)Number of independentvariables(predictors).– Number of predictors should notbe higher than1/3of observatons.
– E.g.For 10mesured firs,3preditors.– Or increase the number of observations (measuremorefirs).
MethodologyJakubHorák
Dataevaluation
(4)Testof predictors for multicollinearity.– Useof varianceinflation factor (VIF).– VIF≥2,avoid predictor.• Stepbystep.• Suddenly.
– Software:SAM,R.
MethodologyJakubHorák
Dataevaluation
(5)First tests – independentvariable iscategorical (dummy).– Comparison of two paired categories.• Paired t-test(non-parametric:Wilcoxon test).
MethodologyJakubHorák
Dataevaluation
(5)First tests – independentvariable iscategorical (dummy).– Comparison of two notpaired categories.• t-test(Mann-Whitney test).
MethodologyJakubHorák
Dataevaluation
(5)First tests – independentvariable iscategorical (dummy).– Comparison of morethan two categories.• ANOVA(Kruskal-Wallis test)+posthoctestamongcategories.
• Sotware:Statistica,R,Excel.
MethodologyJakubHorák
Dataevaluation
(6)First tests – independentvariable iscontinuous.– Linear regression.• Result of the whole model(R2,AIC…andP).• Result of particular variables (tandP).
• Software:Statistica,R,SAM…
MethodologyJakubHorák
Dataevaluation
(7)Advanced modelling.– Generalized linear model.• Possibility of useother than normal distribution.• Combination of categorical andcontinuous variables.• Result of whole model(R2,AIC…aan P).• Result of particular variables (tor zandP).
– Software:Statistica,R,SAM…
MethodologyJakubHorák
Thanks for attention
MethodologyJakubHorák