Hello
I test various materials and I want to use math as a basic quality control in order to verify that I have measured the material in question correctly.
I have arrenged the obtained values from the measurement in a spreadsheet column and sorted them from lowest to highest.
In the box below I have summarized all the values.
In the next column I have calculated the expected values in each row following a perfect normal distribution. So in the first row (first measurement) I have calculated the expected value to be 0,024*sum of obtained values. In the next row (second measurement) I have calculated the value to be 0,159*sum of obtained values. And in the following rows 49,9%, 83,9%, 97,4% and 99,7% of the sum of obtained values.
Then I perform a chi-square test and compare my obtained data with my expected data. If the p-value is <0,05 I reject the hypothesis that my obtained data has a normal distribution and consequently the data is invalid due to man or machine error.
However, the calculation of my expected values is only valid with 6 measurements. What if I have 8, 10 or n measurements? What percentages should I use then? What is the general formula for this and can you find these values in a table somewhere?
I test various materials and I want to use math as a basic quality control in order to verify that I have measured the material in question correctly.
I have arrenged the obtained values from the measurement in a spreadsheet column and sorted them from lowest to highest.
In the box below I have summarized all the values.
In the next column I have calculated the expected values in each row following a perfect normal distribution. So in the first row (first measurement) I have calculated the expected value to be 0,024*sum of obtained values. In the next row (second measurement) I have calculated the value to be 0,159*sum of obtained values. And in the following rows 49,9%, 83,9%, 97,4% and 99,7% of the sum of obtained values.
Then I perform a chi-square test and compare my obtained data with my expected data. If the p-value is <0,05 I reject the hypothesis that my obtained data has a normal distribution and consequently the data is invalid due to man or machine error.
However, the calculation of my expected values is only valid with 6 measurements. What if I have 8, 10 or n measurements? What percentages should I use then? What is the general formula for this and can you find these values in a table somewhere?