turn on suggestions

Auto-suggest helps you quickly narrow down your search results by suggesting possible matches as you type.

Showing results for

- JMP User Community
- :
- Discussions
- :
- Discussions
- :
- Re: Power Analysis Help

Topic Options

- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Get Direct Link
- Email to a Friend
- Report Inappropriate Content

Nov 1, 2018 3:04 AM
(213 views)

Hi,

Please could I get some help understanding power analysis?

I have 2 treatments with 6 pen replicates. I have ran an ANOVA which has showed the P value of 0.1, SEM= 0.011. I would like help to find out:

- With the current data- what is the l.s.d. or the minimum difference between the two points I would need to have to see a significant difference here?
- How many more replicates would I need to see a significant difference between the two mean values I have?

I have seen the power analysis window and I'm guessing I would select:

Solve for Power, Solve for LSV, what does LSN mean? Do I need to enter any numbers into the table for power analysis or can I leave it blank?

In the analysis below, would I be right in assuming that my number of replicates I need would be 14 for each treatment?

And would I need to see a difference of 0.0365 for these two to be significant with the current level of reps?

Thanks in advance

Abbie

1 REPLY

Highlighted
##
##### Re: Power Analysis Help

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Get Direct Link
- Email to a Friend
- Report Inappropriate Content

Hi Abbie,

You can find out more about the interpretation of the power analysis in the oneway analysis here:

https://www.jmp.com/support/help/14/power.shtml#864840

*"LSV (the Least Significant Value) is the value of some parameter or function of parameters that would produce a certain p-value alpha. Said another way, you want to know how small an effect would be declared significant at some p-value alpha. The LSV provides a measuring stick for significance on the scale of the parameter, rather than on a probability scale. It shows how sensitive the design and data are.**LSN (the Least Significant Number) is the total number of observations that would produce a specified p-value alpha given that the data has the same form. The LSN is defined as the number of observations needed to reduce the variance of the estimates enough to achieve a significant result with the given values of alpha, sigma, and delta (the significance level, the standard deviation of the error, and the effect size). If you need more data to achieve significance, the LSN helps tell you how many more. The LSN is the total number of observations that yields approximately 50% power.**Power is the probability of getting significance (p-value < alpha) when a real difference exists between groups. It is a function of the sample size, the effect size, the standard deviation of the error, and the significance level. The power tells you how likely your experiment is to detect a difference (effect size), at a given alpha level.**Note: When there are only two groups in a one-way layout, the LSV computed by the power facility is the same as the least significant difference (LSD) shown in the multiple-comparison tables."*

You might want to look at the sample size and power analysis platform (In JMP14: DOE > Design Diagnostics > Sample Size and Power).

This result agrees with the power analysis in your oneway analysis that to have a 50% chance of seeing a significant difference (p<0.05) when the true difference is 0.02902 or greater and with a std dev of 0.031386 you would need 21 samples in total. Just note that the difference to detect is defined differently in the two platforms: difference from overall mean in Oneway and difference between groups in Sample Size and Power.

50% power may not be enough. 80% is often the required power and that would require 39 samples.

This may seem like a lot but you need to consider that your within group std dev is large compared to the size of difference you want to detect.

I hope that helps.

Phil