Cannon Posted August 19, 2018 Share Posted August 19, 2018 I have been working on adding standard deviation to the ratings data at CPW. Question: What is an acceptable sample size? I've seen stuff as low as 10. I've seen weirdly specific numbers like 26. It seems like everyone agrees 100 is enough. Anyone who is statistics savvy (is this math? chemistry?) have any input? What is an acceptable sample size? 1 Quote Link to comment Share on other sites More sharing options...
richcottrell Posted August 20, 2018 Share Posted August 20, 2018 Great question.... default answer: 2 Quote Link to comment Share on other sites More sharing options...
Cannon Posted August 20, 2018 Author Share Posted August 20, 2018 16 minutes ago, richcottrell said: Great question.... default answer: C’mon Rich! What’s safe? What’s responsible? 1 Quote Link to comment Share on other sites More sharing options...
richcottrell Posted August 20, 2018 Share Posted August 20, 2018 But seriously, you are the only person with real numbers... I got a “D” freshman year at Penn State in basic college algebra so numbers still puzzle me! 1 Quote Link to comment Share on other sites More sharing options...
Tom Flesher Posted August 20, 2018 Share Posted August 20, 2018 Honestly, I think it’s most responsible to report the mean, the SD and the number of observations, but that would require updating somewhat regularly. If you wanted to really get crazy you could report a 90% confidence interval (excel can do this really easily if you’re interested). Off the top of my head I can’t recall a test for whether a standard deviation is statistically reliable. 2 Quote Link to comment Share on other sites More sharing options...
Climber028 Posted August 20, 2018 Share Posted August 20, 2018 I would be more interested in the variance than the standard deviation. As for sample size ironically the amount you should have depends on the deviation, obviously the more the better but I would feel good with 30-50 when speaking about grippers. 1 Quote Link to comment Share on other sites More sharing options...
Cannon Posted August 20, 2018 Author Share Posted August 20, 2018 34 minutes ago, Climber028 said: I would be more interested in the variance than the standard deviation. As for sample size ironically the amount you should have depends on the deviation, obviously the more the better but I would feel good with 30-50 when speaking about grippers. What do you mean by variance. Min and max are already published. Quote Link to comment Share on other sites More sharing options...
Climber028 Posted August 20, 2018 Share Posted August 20, 2018 11 minutes ago, Cannon said: What do you mean by variance. Min and max are already published. You might be thinking of range, which is just the difference between maximum and minimum. Variance is a different value related to standard deviation. Excel calculates it automatically so I won't bore people with the details but in plain language it's a measure of how spread out a set of data is. That's what I would want to know when looking at ratings data, but I'm weird so it might not be what most people need or want. 1 Quote Link to comment Share on other sites More sharing options...
Cannon Posted August 20, 2018 Author Share Posted August 20, 2018 13 minutes ago, Climber028 said: You might be thinking of range, which is just the difference between maximum and minimum. Variance is a different value related to standard deviation. Excel calculates it automatically so I won't bore people with the details but in plain language it's a measure of how spread out a set of data is. That's what I would want to know when looking at ratings data, but I'm weird so it might not be what most people need or want. Okay i’ll check that out. Quote Link to comment Share on other sites More sharing options...
richcottrell Posted August 20, 2018 Share Posted August 20, 2018 You guys have gone over my head, but I thank you for crunching numbers for us... now I will go back to my remedial mathamatical studies: 2 Quote Link to comment Share on other sites More sharing options...
Cannon Posted August 20, 2018 Author Share Posted August 20, 2018 1 hour ago, richcottrell said: You guys have gone over my head, but I thank you for crunching numbers for us... now I will go back to my remedial mathamatical studies: This is all a little over my head too. I don't have any experience using these types of calculations, but can make Excel do it. But even though I can, i'm stopping to wonder if I should. 1 Quote Link to comment Share on other sites More sharing options...
Climber028 Posted August 20, 2018 Share Posted August 20, 2018 It seems like most people are satisfied with the data you provide currently. Tho it's not much extra work to throw in a few extra columns on excel so I can't see a reason not to do it. If that's too cluttered you can even separate it by having your normal ratings data then a secondary section for "advanced analysis" or whatever. Or even sell people the raw dataset. 3 Quote Link to comment Share on other sites More sharing options...
wobbler Posted August 20, 2018 Share Posted August 20, 2018 Graph what you have and add links to the pictures. Max/min/average is enough for most people, the rest can look at the bell curve. The number of people who will fully understand and appreciate the statistical definitions is very small (I'm not one of them.) 2 Quote Link to comment Share on other sites More sharing options...
Tom Flesher Posted August 20, 2018 Share Posted August 20, 2018 3 hours ago, wobbler said: Graph what you have and add links to the pictures. Max/min/average is enough for most people, the rest can look at the bell curve. The number of people who will fully understand and appreciate the statistical definitions is very small (I'm not one of them.) Not a bad idea, considering how easily you can chart this stuff in Excel. If you’re interested in something that looks a bit more professional I could toss up a chart for you in R. It’d take a hot second. 2 Quote Link to comment Share on other sites More sharing options...
Gary Gray Posted August 20, 2018 Share Posted August 20, 2018 6 hours ago, Climber028 said: It seems like most people are satisfied with the data you provide currently. Tho it's not much extra work to throw in a few extra columns on excel so I can't see a reason not to do it. If that's too cluttered you can even separate it by having your normal ratings data then a secondary section for "advanced analysis" or whatever. Or even sell people the raw dataset. I personally like what what Cannon Powerworks is doing now and I am very happy that they are taking the time to do it. No need to change anything. 1 Quote Link to comment Share on other sites More sharing options...
Adam Juncker Posted August 20, 2018 Share Posted August 20, 2018 Standard Deviation would not be that helpful, its just average amount of deviation from the mean. Smaller numbers generally mean a more consistent process, with a SD of 0 being no variation. So.... on grippers that you don't have a large sample size (GHP 10 for example), you would likely see a higher SD and people might be all like "Whoa, them GHP's got some variation" when they really don't, we just don't have enough data. With the information you already provide (Average, Min, Max, and the sample size), those of us that really care can figure out the info we want. Perhaps you could publish the "mode" for each gripper, that would tell us which rating you get the most. It can also help predict a shift in ratings. It might be interesting to break out the ratings by the year too. Everybody knows that grippers are way harder now than they were 5-10 years ago 2 Quote Link to comment Share on other sites More sharing options...
Cannon Posted August 21, 2018 Author Share Posted August 21, 2018 So far I was only taking the time for sample sizes of 100+. Which, is only a handful of the most popular grippers. Quote Link to comment Share on other sites More sharing options...
Cannon Posted August 21, 2018 Author Share Posted August 21, 2018 I appreciate the feedback. Since this is extra work, I likely won’t chase it too hard. Quote Link to comment Share on other sites More sharing options...
Cannon Posted August 21, 2018 Author Share Posted August 21, 2018 7 hours ago, Tom Flesher said: Not a bad idea, considering how easily you can chart this stuff in Excel. If you’re interested in something that looks a bit more professional I could toss up a chart for you in R. It’d take a hot second. What is R? Quote Link to comment Share on other sites More sharing options...
Tom Flesher Posted August 21, 2018 Share Posted August 21, 2018 Just another data management package. I use it on my day job. It’s easy to make nice-looking charts in. Quote Link to comment Share on other sites More sharing options...
Adam Juncker Posted August 21, 2018 Share Posted August 21, 2018 (edited) If your ratings data is already in excel, and your ratings for each gripper are all in one row or column, you can just use the formula to calculate it. =stdev.s(highlight your row/column). This will give you the standard deviation of the sample and it calculates it based on the sample size that you provide. So if you have 200 ratings of CoC3's, then it will calculate it based on the sample size. something like this.... All the gripper ratings here are randomly generated numbers... just for demonstration: Gripper Ratings Data Coc 3 GHP 7 157 146 150 147 153 152 146 149 150 143 152 142 138 154 164 151 150 144 146 146 166 142 153 142 163 150 143 152 144 143 160 154 Min 138 142 formula for min is "=min(B2:B19)" Max 166 154 Formula for max is "=max(B2:B19)" Mean 151.8 147.1 Formula for mean is "=average(B2:B19)" Sample size 18 14 Formula for sample size is "=count("B2:B19") Standard Dev. 7.90 4.20 Formula for standard deviation is "=stdev.s(B2:B19)" Edited August 21, 2018 by Adam Juncker Quote Link to comment Share on other sites More sharing options...
Cannon Posted August 21, 2018 Author Share Posted August 21, 2018 I am using array formulas because everything is in a huge list by order # and the formula has to check multiple points of data from different columns to pull out min/max, etc. Quote Link to comment Share on other sites More sharing options...
Recommended Posts
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.