Data Trend Analysis
Six Sigma – iSixSigma › Forums › Old Forums › General › Data Trend Analysis
 This topic has 4 replies, 5 voices, and was last updated 14 years, 11 months ago by AlexanderSO.

AuthorPosts

October 17, 2007 at 2:55 pm #48439
I have four years of data showing job shorts off a manufacturing line. My goal is to determine if there is significant difference between years or if the trends are similiar. I graphed the four years of data in Excel and they overlap fairly well, is there a test which tells me how well the four years fit?
0October 17, 2007 at 3:50 pm #163269
Donotknow6sigmaParticipant@Donotknow6sigma Include @Donotknow6sigma in your post and this person will
be notified via email.Perform Anova test then do Tukey Test in case you reject the Null.
0October 17, 2007 at 4:30 pm #163271You have a variety of options, using run charts or control charts to assess trends, scatter plots for correlation, box plots for comparison analysis, paretos for significance, histograms for characterizations, etc.
The Anova that was earlier suggested might be an option if the underlying assumptions are met, but I would be a bit cautious with that approach.
You are dealing with happenstance data that is over four years old and gathered using an unknown method and untested measurement system (I assume). I might use the data to get some general ideas for future, more specific efforts, but I dont think I would jump into any direct reports with that data.
Good luck.0October 17, 2007 at 5:49 pm #163276I agree with annon and will only add that regression is another option if all the correlation, scatter plot assumptions are met. also depends on the data type you have – if your comparing time series data you could do this by comparing different years variables.
0October 29, 2007 at 2:46 pm #164085
AlexanderSOParticipant@AlexanderSO Include @AlexanderSO in your post and this person will
be notified via email.Hi, if your main focus is to determine if the various set of data fits one to each other you can use a two step method.
The first step is to check the difference between the series usig min square error, for each data calculate as follows (reference valueevaluated value)^2 then at the end add all the values and divide them by the number of data evaluated and this number represents the difference between the series.
The second step is visual, you can plot the series one over each other and check if the behavior is the same.0 
AuthorPosts
The forum ‘General’ is closed to new topics and replies.