Measuring significant change between single unrelated variables?

Six Sigma – iSixSigma Forums Operations Marketing and Sales Measuring significant change between single unrelated variables?

Viewing 2 posts - 1 through 2 (of 2 total)
  • Author
  • #255272


    What are some tests which could alert you that a specific number like a sales figure or changes in % (ratio), measured at different intervals like daily or weekly, are changing in a way that is statistically significant from it previous measurement(s) of that same single figure? Perhaps the two samples tests can work in that sample 1 would be the last measurement and sample 2 would be the newest measurement?

    I can’t seem to find a good answer and everything I research focuses on differences between two groups rather than just significant change with a single variable.

    Here are some example of variables that people may want to measure in business on a daily or weekly basis, to give some context:
    * number of customers serviced
    * costs
    * income
    * on-time delivery % (so % of packages delivered that were on time)

    If we service 20 more customers today than we did yesterday is that significant? If we spend or perhaps make $5000 more dollars this week than we did last week is that significant? If I have a measurement that is 6% this month and 9% next month was that a significant?

    For all these above questions I would also be curious if there is a way to tell at which point it’s significant. So if $5,000 is a significant change in revenue from week to week how would I also determine a threshold like “any increase or decrease of $1,200 or more is considered significant.” which would also explain why $5,000 is significant, it’s above the $1,200 threshold.

    Any insight someone would lend would be much appreciated.


    Robert Butler

    The short answer to your question is there aren’t any tests like that. In order to make a meaningful comparison of any kind you need context. Comparing one single number with another single number is a comparison devoid of context. The only thing you can say about any observed difference between two single numbers is a difference exists.

    The way you get context is to collect data on whatever process you have. The data will give you a sample size and analysis of the data will give you an estimate of typical and the ordinary process variation around typical. Once you have these things you use this information to make assessments concerning differences – for example:

    1. You can ask if there is a significant difference between what the process has been doing and what your most recent measurement tells you.

    2. You can use the estimate of ordinary variation of the population to identify the minimum size of a difference needed before one could claim a statistically significant difference between any two single measurements.

Viewing 2 posts - 1 through 2 (of 2 total)

You must be logged in to reply to this topic.