Structure of the data:
- Nominal data: gender, occupation, marital status, religious beliefs and so on
- Order data: education, job title, hospital level
Numerical data divided into two categories
- Count data
- Metering data
Descriptive qualitative data analysis ( "single categorical variable")
- Chart method (bar chart, FIG consistent (produced just 18 years), and the like)
- Numerical calculation (the ratio of the relative risks and odds ratios)
Associations of qualitative variables ( "multiple categories Study variables")
- Contingency Analysis: Two the two study, a time study of the relationship between two variables
- Correspondence Analysis: "variable relationship," "ordering relations", "polynomial regression (generalized)"
- High dimensional column associated with the best Pictured: "FIG mosaic"
- FIG consistent (the amount of information provided by a little more), when the relationship between the variables corresponding sample studied may be consistent with FIG.
Front diagram method, the following numerical method is
the same if the samples taken and the overall structure of the sample, the percentage of the ranks can easily use, or can not easily use
Fit Test: Verify that a thing is right
Set the null hypothesis: The default is usually the case, most of the cases, the main case, it is assumed that rarely occurs as a backup
Correspondence analysis and regression analysis
The correlation coefficient:
$\frac{\sum{(x_i-xjz)(yi-yjz)}}{\delta_x*\delta_y}$
insert here the picture description
1. Composition analysis points
2. Canonical Correlation Analysis
Analysis of the code corresponding to R
library(MASS)
data("Suicide")
A1<-xtabs(Freq~age.group+method2,subset=sex=="female",data=Suicide)
A2<-xtabs(Freq~+age.group+method2,subset=sex=="male",data=Suicide)
AA1<-matrix(A1,ncol=8)
AA2<-matrix(A2,ncol=8)
rownames(AA1)=c("F15","F30","F45","F60","F80")
colnames(AA2)=c("poison","gas","hang","drown","gun","knife","jump","other")
rownames(AA2)=c("M15","M30","M45","M60","M80")
colnames(AA2)=c("poison","gas","hang","drown","gun","knife","jump","other")
A<- rbind(AA1,AA2)
ca<-corresp(A,nf=2)
plot(ca)
abline(v=0,h=0,lty=5)
Regression analysis of qualitative data
What kind of return can do it?
Simpson Refutation
extend:
== "statistical world" == Purdue University
Statistical data collection, statistical science and art
Percent favorite bar chart in which statistics
R package with two languages
vcd
mass