X-Message-Number: 27761 References: <> From: David Stodolsky <> Subject: Re: FDGD 2006 Survey Analysis/Commentary (LONG) Date: Sun, 26 Mar 2006 12:26:16 +0200 On 23 Mar 2006, at 14:40, Kennita Watson wrote: > Thanks for the input, David! One note first: I'd > be interested to find out if there are any questions > on which there are major (even if not significant) > differences between the answers given by men and by > women, or by people in different age ranges. Consider x Age Enjoy x Sex Future x Sex Accessible x Sex Max Age x Sex were tried, but showed nothing. I didn't adjust the significance levels to account for multiple tests, so we already risk overworking the data with the first round of results. > >> From: David Stodolsky <> >> Subject: Re: FDGD 2006 Survey Analysis/Commentary (LONG) >> Date: Wed, 22 Mar 2006 20:06:53 +0100 >> >> On 20 Mar 2006, at 22:37, Kennita Watson wrote: >>> Frozen Dead Guy Days 2006 >>> Cryonics Awareness/Attitudes Survey >>> >>> Questions About Grandpa Bredo >>> 1) Do you think Grandpa Bredo has been properly cryopreserved? 12 >>> _Yes 30 _No 38 _Don't know >> >> Not Significant (NS) result. >> >> (A non-significant result means that the data can be accounted for by >> random responses, data getting scrambled in transfer to machine >> readable format, etc. However, this doesn't mean that data were >> scrambled, etc.) > > It seems to me that random responses would > give a much more even distribution. 30 and 38 > are pretty close, but with 12, it seems clear > that fewer people think Yes than No. For the > Don't Knows to even things out, they'd have to > be biased almost 3 to 1 toward Yes. What am I > missing here? We are not testing the difference between "Yes" and "No", but the difference among the 3 possible choices. I made an error in reading off the results for the distributions (those tests related to a single variable). I'll repost with the correct results. >> >>> What I was looking for was the spread between this and the >>> answers to >>> Q11-Q13, which demonstrate that people see the difference in the >>> chances afforded by modern cryonics procedures. >> >> Scales are different, making comparisons of limited value: >> >> Mean difference is 1.9 or 19%, Highly significant (p<.0001). So, >> people can see the difference. > > Does the number mean that the mean chance given in Q13 > is 47%? No. This tests the difference between the two variables (the probability that) science (will make Bredo's revival possible) and (how many years it will take to achieve revivable) preservation. >> >>> 3) If >0%, how many years do you think it will be before he is >>> revived? >>> 4 _<20 >>> 16 _20-50 >>> 15 _50-100 >>> 13 _100-200 >>> 1 _200-500 >>> 2 _>500 >>> Two people who said 0% to Q2 answered Q3 too (they were the only >>> ">500" people). I think I could combine the two and alleviate >>> confusion by having a "Never" choice on Q3. Basically, people >>> figure >>> that if we can revive him at all, we can do it in the next 200 >>> years. >> >> Failure to use an interval scale makes it impossible to test this >> without grouping the data. >> Post-hoc grouping eliminates any valid test options. > > I thought that what I used *was* an interval scale. An interval scale has *equal* intervals. You have intervals of 30yrs, 50yrs, etc. > The interval scale is what did the grouping. Or > maybe "post hoc" means something different than I > thought. Combining 200-500 and 500+ is creating a group after you collected the data (post-hoc). > I'm also not sure what test options you > think were eliminated by the post-hoc grouping I > don't think I was doing. Whether people figure that it can be done within the next 200 years is what could have been tested - if you had decided in advance that is what you wanted to know. > >> The best that can be said is that random responding can be rejected >> as an explanation for the last two values. >>> >>> Questions About Frozen Dead Guy Days >>> 5) The Frozen Dead Guy Days trivialize cryonics and make it look >>> ridiculous. 2 Yes / 56 No / 22 Maybe >>> It's good to see that No outnumbers Yes and Maybe together by more >>> than 2 to 1. >> >> This is post-hoc grouping. > > Why is that bad here? You are combining Yes and Maybe after the seeing the data. >> >> >>> 12) If so, how many years would you guess it will be before the >>> first >>> successful revival? >>> 6 _<20 >>> 23 _20-50 >>> 20 _50-100 >>> 15 _100-200 >>> 3 _200-500 >>> 3 _>500 >> >> Sig. result. > > I think you're saying that the result is > not due to chance with some high probability. > Can you say in English what that "result" is? There isn't much more that can be said after looking at the data. With these intervals, all we can say is that most chose 20-50 yrs. >> >> >>> 14) Before Frozen Dead Guy Days 2006, did you know that cryonics is >>> accessible to the general >>> public (for example, payable by means of a life insurance policy)? >>> 36 _Yes 43 _No >> >> NS > > What this says to me in English is that about > the same number of people knew as didn't know. > The difference is not significant, and the > lack of difference is significant. Something > like that. No. A non-significant result tells you nothing, except that it could be due to chance. > >>> Questions About You >>> 16) What is your current age? >>> 7 _<18 >>> 11 _18-25 >>> 18 _26-35 >>> 17 _36-45 >>> 16 _46-55 >>> 9 _56-65 >>> 2 _66-75 >>> 0 _>75 >> >> Sig. > > Significantly what? There were differences in ages. > Does it give an average > age? No. Equal intervals are needed for that. Better to just let them write their age, that gives a ratio scale, which is even better. >> >>> 18) How much would you say you enjoy life overall now? >>> Not at all >>> some >>> pretty well >>> very much! >>> 0 1 2 3 4 5 6 7 8 >>> 9 10 >> >> Median = 9 >> Mean = 8.3 > > How much would those numbers have to change > (with a similar sample size) to say that the > change was significant? Depends upon the variance in the responses. >> >>> 19) Till what age would you like to live if you could be in good >>> health the entire time? (circle one) >> >> Sig. > > Significantly what? Choices were different among the persons responding. >> >>> 20) What do you guess the world will be like in the far future, >>> compared to today? >> >> NS >> > Seems here to just mean "pretty wide spread". No. A uniform distribution can be significant. > Again it seems that getting a mean and median > so they could be compared to other surveys > would be instructive. You can't get a mean from ordinal data. >>> This graph cross-correlating the answers to Q20 and Q21 corroborates >>> the association between optimism about the future and willingness to >>> consider cryonics. >> >> There is a trend here, but there isn't enough data to safely say the >> result is valid. > > How much data would it take? Depends upon the data. > And how is "valid" > different from "significant"? Or is there a > significant trend? Of three variations of the same test, one was significant at the .05 level. However, there was a warning that the result might not be valid, because the expected number of data points in over 20% of the cells in the table was below 5. >> >> I have not performed any test of the differences between 2005 and >> 2006 data, since the data collection procedure was different. This >> makes it impossible to say whether a change results from a different >> procedure or a difference in the population. > > How was the data collection procedure different in cases > where the questions were the same? Questions can influence each other's results. The potential reward, etc. was different, dss David Stodolsky Skype: davidstodolsky Rate This Message: http://www.cryonet.org/cgi-bin/rate.cgi?msg=27761