CODEWITHSUNDEEP
rgaussian
Mia
Mia

Reputation: 1

how can i draw a gaussian curve on datas which obey gaussian distribution by R language?

I have some datas which looks like obeying gausssian distribution. So i use

my.glm<- glm(b1~a1,family=Gaussian)

and then use command

summary(my.glm).

The results are:

Call:
glm(formula = b1 ~ a1, family = gaussian)

Deviance Residuals: 
      Min         1Q     Median         3Q        Max  
-0.067556  -0.029598   0.002121   0.030980   0.044499  

Coefficients:
             Estimate Std. Error t value Pr(>|t|)    
(Intercept)  0.433697   0.018629   23.28 1.36e-12 ***
a1          -0.027146   0.001927  -14.09 1.16e-09 ***
---
Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1 

(Dispersion parameter for gaussian family taken to be 0.001262014)

Null deviance: 0.268224  on 15  degrees of freedom
Residual deviance: 0.017668  on 14  degrees of freedom
AIC: -57.531

Number of Fisher Scoring iterations: 2

I think they fit well. But how can i draw a gaussian curve on these datas?

Upvotes: 0

Views: 1316

Answers (2)

David LeBauer
David LeBauer

Reputation: 31761

Assuming that the intercept has a normal distribution, you can plot its distribution like this:

x <- seq(0.3,0.6,by =0.001)
plot(x, dnorm(x, 0.433697, 0.018629), type = 'l')

and you might want to add your data:

rug(b1)

since you didn't supply data, we can make some up (with some transforms to match stats in the example):

set.seed(0)
b <- rnorm(15)
b1 <- ((b - mean(b))/sd(b)  * 0.018629) + 0.433697
rug(b1)

you could also overlay a kernel density estimate of the data

lines(density(b1), col = 'red')

Giving the following plot:

enter image description here

Upvotes: 1

Carl Witthoft
Carl Witthoft

Reputation: 21532

Simple: ?dnorm

Use dnorm to create a gaussian curve of desired mean and s.d. without tying yourself to any numerically fitted function. This is a simple, and good, way to show how your data 'fits' to a theoretical curve. Not the same thing as plotting the fitted data and trying to figure out "how close" to a gaussian it is.

Upvotes: 0

Related Questions