Problem of sample size in control charts

Hi,

I am a newbie in statistical process control (SPC). I would like to do a visual comparison between xbar-s, CUSUM and EWMA charts. I would like to use similar datasets for this:

Question 1 (see datasetSingle.csv)

A single measurement (sample size n=1) is available and taken every 20 seconds for the weight of an item from a production process. There are 231 readings available. How can I use this dataset to construct xbar-s, CUSUM and EWMA charts in MATLAB so that I can compare their sensitivities? (see datasetSingle.csv)

Question 2 (see datasetMultiple.csv)

Three measurements (sample size n=3) are available and taken every minute for the weight of an item from a production process. There are 25 readings available. How can I use this dataset to construct xbar-s, CUSUM and EWMA charts in MATLAB so that I can compare their sensitivities? (see datasetMultiple.csv)

Answer 1

[Rahul]

Hi @Andy,

In order to achieve the required control charts of types xbar-s, CUSUM and EWMA , MATLAB provides 'controlchart' and 'cusum' functions.

The 'controlchart' function contains a 'charttype' property which can be set to 'xbar', 'ewma', 's' and several other types as well.

Here is an example:

data = csvread('datasetSingle.csv'); % Reading data 
controlchart(data, 'charttype', 'ewma'); % Chart of type ewma

Note: Chart of type 'xbar' would require grouped data.

Additionally, 'cusum' function can be used to obtain the chart with cumulative sum. Here is an example:

cusum(data); % Chart with cumulative sum

For the dataset with multple redings, considering a mean of the values using the 'mean' function or some other method would be viable to obtain consistent data to be used with 'controlchart' and 'cusum' functions.

The following MATLAB Answer can be referred:

https://www.mathworks.com/matlabcentral/answers/1601885-help-creating-xbar-control-chart

The following MathWorks documentation can be referred to knwo more:

'controlchart': https://www-test1.mathworks.com/help/releases/R2018b/stats/controlchart.html

'cusum': https://www-test1.mathworks.com/help/releases/R2018b/signal/ref/cusum.html

Hope this helps! Thanks.