the accuracy of the device

[Omar Farook]

Dear Experts,
  long time have been passed since my last inquiry , I hope all of you in  best-situation.
  My question today is.
  My colleague would  like to examine accuracy or the reliability of a  device , the device measuring teeth's  strength.
  He has 50 teeth , he will measure the teeth’s strength  today and after 2 or 3 days he will do the same process by using the same device , to examine the accuracy of the device.   Could we use Paired-Samples T Test if the data follow the normal distribution or we should use Two-Related-Samples Tests if the data non normal?  Many thanks.
  Omar.


---------------------------------
How low will we go? Check out Yahoo! Messenger’s low  PC-to-Phone call rates.

[Marta García-Granero]

Hi Omar

OF>   My colleague would  like to examine accuracy or the
OF> reliability of a  device , the device measuring teeth's  strength.
OF>   He has 50 teeth , he will measure the teeth’s strength
OF> today and after 2 or 3 days he will do the same process by using
OF> the same device , to examine the accuracy of the device.   Could
OF> we use Paired-Samples T Test if the data follow the normal
OF> distribution or we should use Two-Related-Samples Tests if the
OF> data non normal?

Neither test you mention will answer your question (measurement
error), since both focus on whether there is a bias (systematic
difference between the two days). I suppose you want statistics to
verify the repeatability of a device (the ammount of random
variation). A p-value (obtained by any test) will not answer the
question. Take a look at these links:

http://bmj.bmjjournals.com/cgi/content/full/313/7059/744
http://bmj.bmjjournals.com/cgi/content/full/313/7048/41
http://www-users.york.ac.uk/~mb55/meas/

Time ago I wrote some code for the methods discussed in those appers.
Let me know if you are interested in having it.




--
Regards,
Dr. Marta García-Granero,PhD           mailto:[hidden email]
Statistician

---
"It is unwise to use a statistical procedure whose use one does
not understand. SPSS syntax guide cannot supply this knowledge, and it
is certainly no substitute for the basic understanding of statistics
and statistical thinking that is essential for the wise choice of
methods and the correct interpretation of their results".

(Adapted from WinPepi manual - I'm sure Joe Abrahmson will not mind)

[Omar Farook]

Dear Dr. Marta,
  Many thanks for the prompt help.
  I have certain questions with respect to article 1 , are the standard deviation of repeated measurements on the same subject is the measurement error of that subject?
  What is the meaning of the following phrase?
  “measurement error depends on the size of the measurement”
  And finally , if there is no relation between the sd and the mean , could we said that the device is reliable?
  Many thanks in advance.
  Omar.


Marta García-Granero <[hidden email]> wrote:  Hi Omar

OF> My colleague would like to examine accuracy or the
OF> reliability of a device , the device measuring teeth's strength.
OF> He has 50 teeth , he will measure the teeth’s strength
OF> today and after 2 or 3 days he will do the same process by using
OF> the same device , to examine the accuracy of the device. Could
OF> we use Paired-Samples T Test if the data follow the normal
OF> distribution or we should use Two-Related-Samples Tests if the
OF> data non normal?

Neither test you mention will answer your question (measurement
error), since both focus on whether there is a bias (systematic
difference between the two days). I suppose you want statistics to
verify the repeatability of a device (the ammount of random
variation). A p-value (obtained by any test) will not answer the
question. Take a look at these links:

http://bmj.bmjjournals.com/cgi/content/full/313/7059/744
http://bmj.bmjjournals.com/cgi/content/full/313/7048/41
http://www-users.york.ac.uk/~mb55/meas/

Time ago I wrote some code for the methods discussed in those appers.
Let me know if you are interested in having it.




--
Regards,
Dr. Marta García-Granero,PhD mailto:[hidden email]
Statistician

---
"It is unwise to use a statistical procedure whose use one does
not understand. SPSS syntax guide cannot supply this knowledge, and it
is certainly no substitute for the basic understanding of statistics
and statistical thinking that is essential for the wise choice of
methods and the correct interpretation of their results".

(Adapted from WinPepi manual - I'm sure Joe Abrahmson will not mind)



---------------------------------
Yahoo! Music Unlimited - Access over 1 million songs.Try it free.

[Marta García-Granero]

Hi Omar

OF>   I have certain questions with respect to article 1 , are
OF> the standard deviation of repeated measurements on the same
OF> subject is the measurement error of that subject?

Sort of. More interesting than the SD is the "repeatability"
(2.77*SD), because it gives the average difference you could find
between two consecutive measures of the same subjects 95% of the
times. If this quantity is considered small enough, then you can
consider that the measurement device is accurate for your purposes.

OF>   What is the meaning of the following phrase?
OF>   “measurement error depends on the size of the measurement”

Sometimes the degree of heterogeneity is bigger when measures high
values.

OF>   And finally , if there is no relation between the sd and
OF> the mean , could we said that the device is reliable?

No. The analysis that Bland&Altman proposes is based on the assumption
that the SD of the measurement is independent of its mean value.
Therefore, it is important to test this, but it doesn't mean than the
device is reliable (see point number one to answer than question).





--
Regards,
Dr. Marta García-Granero,PhD           mailto:[hidden email]
Statistician

---
"It is unwise to use a statistical procedure whose use one does
not understand. SPSS syntax guide cannot supply this knowledge, and it
is certainly no substitute for the basic understanding of statistics
and statistical thinking that is essential for the wise choice of
methods and the correct interpretation of their results".

(Adapted from WinPepi manual - I'm sure Joe Abrahmson will not mind)