how to test a hypothesis with t-test or p value

Question

farfar 2017-3-22

1
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/331354-how-to-test-a-hypothesis-with-t-test-or-p-value

评论： farfar 2017-3-22

Hello I am testing the null hypothesis mu=3.4 against mu>3.4, for a sample of size 9. I want to test this with both t-distribution table and p-value but its giving me NaN instead of 0 or 1.this is the code im using :

mu = 3.4;     % Population mean
sample=[3.4,3.6,3.8, 3.3,3.4,3.5,3.7,3.6,3.7]
n = numel(sample)
xobs = mean(sample)  % Sample mean
s = std(sample)      % Sample standard deviation
t = (xobs - mu)/(s/sqrt(n))
p = 1-tcdf(t,n-1)
[h,ptest] = ttest(xobs,mu,0.05,'right')

Thank you

0 个评论
显示 -2更早的评论隐藏 -2更早的评论

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

the cyclist 2017-3-22

3
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/331354-how-to-test-a-hypothesis-with-t-test-or-p-value#answer_259914

在 MATLAB Online 中打开

I can't look at this fully right now, but one thing I notice immediately is that you are putting the mean value of the sample as the first argument of ttest(), but you should be putting the sample itself:

ttest(sample,mu,0.05,'right')

I don't know if that is the only problem with your code, but is is definitely a problem with it. :-)

4 个评论
显示 2更早的评论隐藏 2更早的评论

the cyclist 2017-3-22

编辑：the cyclist 2017-3-22

在 MATLAB Online 中打开

You can't do a t-test with just the sample mean. You need the distribution. This is pretty obvious if you consider the two samples

sample = [10 10 10 10 10 10 10];

and

sample = [-8 0 7 10 13 20 28];

which have the same mean, but obviously different likelihood of being different from the population mean.

The specific reason you get NaN from what you are doing is that you are putting in a "sample" of one value, and the t-test is ill-defined for a sample size of one.

farfar 2017-3-22