problem on simulation code (the loop unable to function effectively)

Mon, Apr 18, 2016 6:25 PM

Hi Si Jie,
Again, please send questions to the list, not me.

Okay, I may have worked out what you are doing. The program runs and
produces what I would expect in the rightmost columns of the result
"g".

You are storing the number of each test for which the p value is less
than 0.05. It looks to me as though the objects storing the results
should be vectors as you are only storing 100 p values at a time.
Using matrices in which the only part of the values are used is a bit
confusing. Perhaps this resulted from my initial suggestion of using
matrices and storing all of the results before calculating the number
of p values less than 0.05, which you didn't do.

If I rewrite your code with vectors for storing the p values, I get
the same results. Essentially, with the scale parameter set the same
for both distributions, you get about five false positives (Type I
errors) per 100 simulations. This is expected. As the scale
parameters, and thus the means, of the groups diverge, you first get a
large number of results less than 0.05. With large differences between
the means, all results are less than 0.05, which seems correct to me.
You have probably been asked to interpret the effect of group size and
difference in means on the number of p values less than 0.05. If the
difference in means is large enough, you are almost guaranteed this.
In the cases where this is not occurring (i.e. the first difference in
scale parameters) you can see the difference in outcomes between the
three significance tests.

Jim

On Tue, Apr 19, 2016 at 10:35 AM, <uk31429 at student.umt.edu.my> wrote:

Hi, i am sorry but that part is not a part of the code.
I am just listing my variable for simulation out at the beginning in order
to provide easy understanding.
i am doing a whole night to find out where have i done wrong.But i just
keep repeat producing the wrong output ...

Hi Jeem,
First, please send questions like this to the help list, not me.

I assume that you are in a similar position to sjtan who has been
sending almost exactly the same questions.

The problem is not in the loops (which look rather familiar to me) but
in your initial assignments at the top. For instance:

scale parameter=(1,1.5,2,2.5,3)

produces an error which has nothing to do with the loops. This is a
very basic mistake, for:

scale_parameter<-c(1,1.5,2,2.5,3)

fixes it. I think if you learn a bit about basic R coding you will be
able to fix these problems yourself.

Jim


On Tue, Apr 19, 2016 at 4:03 AM,  <uk31429 at student.umt.edu.my> wrote:

Greeting dr jim,

I am student from Malaysia. I am doing R simulation study. For your
information, I have been written a code relating to 2 gamma distribution
with equal skewness.
skewness=1.0
shape parameter=16/9
scale parameter=(1,1.5,2,2.5,3)

Below are my coding, however, the code have some error and yet after
trying this and that for a whole day, i couldn't spot the mistake. The
output should be greater or smaller than 0.05 while no exceeding too
much
.But the for loop only able to function for scale parameter 1 .I try to
apply another for loop for the scale parameter, but the output become
worsen, as the simulation will even get hang. all of the output is 100.
Please, could you give me some advice ?

#For gamma disribution with equal skewness 1.5
rm(list=ls())
nSims<-100
alpha<-0.05

#here we declare the random seed generator
set.seed(3)

## Put the samples sizes into matrix then use a loop for sample sizes
sample_sizes<-matrix(c(10,10,10,25,25,25,25,50,25,100,50,25,50,100,100,25,100,100),nrow=2)

#shape parameter for both gamma distribution for equal skewness
shp<-rep(16/9,each=45)

#scale parameter for sample 1
#scale paramter for sample 2 set as constant 1
d1<-matrix(c(1,1.5,2,2.5,3),ncol=1)
scp<-rep(d1,9)

#create a matrix combining the forty five cases of combination of sample
sizes,shape and scale parameter
all<- cbind(rep(sample_sizes[1,],5),rep(sample_sizes[2,],5),scp)

# name the column samples 1 and 2 and standard deviation
colnames(all) <- c("m", "n","scp")


#set empty vector of length to store p-value
equal3<-rep(0,nrow(all))
unequal4<-rep(0,nrow(all))
mann5<-rep(0,nrow(all))

#set nrow =nsims because wan storing every p-value simulated
#for gamma distribution with equal skewness
matrix3_equal  <-matrix(0,nrow=nSims,ncol=3)
matrix4_unequal<-matrix(0,nrow=nSims,ncol=3)
matrix5_mann   <-matrix(0,nrow=nSims,ncol=3)


# this loop steps through the all_combine matrix
 for(ss in 1:nrow(all))

  {  #generate samples from the first column and second column
     m<-all[ss,1]
     n<-all[ss,2]

      for ( sim in 1:nSims)
     {
        #generate 2 random samples from gamma distribution with equal
skewness
        gamma1<-rgamma(m,16/9,all[ss,3])
        gamma2<-rgamma(n,16/9,1)
    #minus population mean from each sample to maintain the equality of
null    #hypotheses (population mean =scale parameter *shape
parameter)
gamma1<-gamma1-16/9*all[ss,3]
gamma2<-gamma2-16/9

     matrix3_equal[sim,1]<-t.test(gamma1,gamma2,var.equal=TRUE)$p.value
     matrix4_unequal[sim,2]<-t.test(gamma1,gamma2,var.equal=FALSE)$p.value
     matrix5_mann[sim,3] <-wilcox.test(gamma1,gamma2)$p.value

      }
##store the result
equal3[ss]<- sum(matrix3_equal[,1]<alpha)
   unequal4[ss]<-sum(matrix4_unequal[,2]<alpha)
      mann5[ss]<- sum(matrix5_mann[,3]<alpha)
}
g<-cbind(all, equal3, unequal4, mann5)

I will be really appreciated for any respond .Thanks.

Your sincerely
Jeem




Computational Mathematics
School of Informatics and Applied Mathematics
Universiti Malaysia Terengganu
21030 Kuala Terengganu, Terengganu, Malaysia
Phone : 017-7799039
Email : uk31429 at student.umt.edu.my


Tan Si Jie UK31429
Computational Mathematics
School of Informatics and Applied Mathematics
Universiti Malaysia Terengganu
21030 Kuala Terengganu, Terengganu, Malaysia
Phone : 017-7799039
Email : uk31429 at student.umt.edu.my