December 21, 2005 at 10:22 am #41790
Theoretically, we are assuming the data is random and following normal distribution before we go for statistic test such as 2 sample t, 1 sample Z. The test is normally tested with 5 % error (95% confidence level).
However, in real case, we will test our data for randomness (5% error) and normality (5% error), then only we will go for statistic test with 95 confidence level (5% error).
So, my question is that the result we obtain after statistic test still having only 5% error? Or it should be taken account for the error during test for randomness and normality as well? If this case happen, the error of our final result will me much more higher than 5%. How can we calculate the total error?
Please advise0December 21, 2005 at 10:57 am #131447
All these tests (Run, Normality & T ) are independent of each other. So there is no such thing called total error.0December 22, 2005 at 1:57 am #131480
You mentioned that all the tests are independent.
However, if we have the 0.95 probability that our randomness test correct, 0.95 probability that the normality test correct, 0.95 probability that equal varaince test correct and 0.95 2-sample t test correct.
Then, should we say that we are making a final conclusion on probability (0.95)^4 = 0.81, is it correct?
The forum ‘General’ is closed to new topics and replies.