With 5 groups and 9 factors: what question(s) are you trying to answer? Depending on the answer to this you may be better off with a regression method.
Anyway ANOVA assumes the variances between the groups are equal. Therefore, this assumption requires validation. If the variances are not equal there are non-parametric and regression methods that…[Read more]
The first thing that may be helpful it to specifically formulate questions that you want to answer. For example, is there a statistically signficant difference in the proportion of missed medications among floors?
From you description of the data it looks as though a Chi-Square Test can be used to answer some the questions. And if your…[Read more]
The p-value represents the probability of rejecting the null hypothesis when it is true. As the analyst you can determine if this is a level of risk you are willing to live with. Traditionally p-values <.05 is considered a sufficient level of risk to reject the null hypothesis.
A good DMAIC Six Sigma project is a problem where you don’t readily know it’s solution. I both of the potential projects you have listed the rist question is: what is the problem? And the follow up question is: do I know the solution? If the solution is known then you don’t need to apply Six Sigma.
Using a CT Flowdown is a good way of f…[Read more]
If all you are intersted in is determining whehter one time period is signficantly from another time period there are two tests you can use: two proportion test and the chi square. However, if what you want to know is whether the process is stable or unstable then a p-chart or IMR chart can be used.Hope this helps.Bob
There are various types of linear regression analysis that can be performed. The key to using them correctly is to assess the regression diagnostics after the analysis is run in order to ensure that the assumptions of the analysis are met. If the assumptions are not met then the output from the analysis can be questionable. If you are not that…[Read more]
I think combining severity and process together will provide a more complete picture. With severity alone the analysis will not reveal the impact of processes and with processes alone the patient’s severity of illness is not considered. However, if you have the time conducting all three analysis could be insightful as well.
There are two generic factors that impact a patient’s length of stay: severity of illness and processes. In order to discern which of the factors has a statistically signficant impact on LOS a multivariate regression analysis would be helfpul. Severity of illness variables include: age, comorbid conditions, lab values, etc. Process variables i…[Read more]
Before you select as sampling strategy and design you need to know what type(s) of statistical analysis you are going to conduct. That is, what question(s) are you trying to answer. From there you can begin to select the appropriate strategy.
Thanks for the reply. In employing the Cpk I’m using continuous data not discrete. If you are sampling data and deriving process capability indices I think it’s important to ascertain whether the indices have significantly changed over time or after an improvement interventioned has been implemented. Just comparing two indices wit…[Read more]
Thanks for the reply. What I’m looking for is an index such as Cp or Cpk for binomial data. Since binomial data approximates the normal distribution and the standard deviation is derived as the square root of npq, it seems plausible that these indices can be reproduced for binomial data. So I’m searching for references that addresses this.
It’s probably not a serious problem from the perspective of the control chart and how limits are derived. It does sound like a serious problem from a measurement system perspective. It sounds like they need to perform a gage R&R to ascertain whether there is a measurement problem.
Before using either the c or u chart you need to ascertain whether the data conforms to a poisson distribution. The control limits on these charts assume the data is distributed as poisson. This is easily accomplished with most statistical packages. Use negative binomial regression with just the dependent variable and assess whether the al…[Read more]
Wheeler is correct if you are plotting mean values on an IR chart, hence, the central limit is invoked. However, when plotting census data – unless you are plotting mean census – on an IR chart the data must be normally distributed. See references other that Wheeler about IR charts.
Unable to tell from the information provided. Is the data normally distributed? What type of control chart are you using? Are there signs of special cause variation? More importantly; if special cause variation is present what will you be able to do about it? Unless you have stratified hospital admissions by type of admission it will be very…[Read more]
- Load More