|
From: | Alan Mead |
Subject: | Re: Bootstrapping on single variable |
Date: | Wed, 12 Dec 2018 12:41:08 -0600 |
User-agent: | Mozilla/5.0 (Windows NT 6.1; WOW64; rv:60.0) Gecko/20100101 Thunderbird/60.3.2 |
I don't think so. I think bootstrapping is a relatively new addition to SPSS that hasn't made it into PSPP. You want FREQUENCIES to have a "Bootstrap" button: that brings up a dialog like this: Which would generate syntax like this: BOOTSTRAP /SAMPLING METHOD=SIMPLE /VARIABLES INPUT=churn /CRITERIA CILEVEL=95 CITYPE=PERCENTILE NSAMPLES=1000 /MISSING USERMISSING=EXCLUDE. FREQUENCIES VARIABLES=churn /STATISTICS=MEAN /ORDER=ANALYSIS. And the output would include bootstrapped estimates of a 95%CI for the mean and for each percent in the frequencies table, like this: Is that correct? For example, I collected a small sample of data (N=112) on some items, and about 3% of the sample flagged item 007 as problematic. The bootstrap option gives me a bootstrapped 95%CI for the population number to be [0%,6%]: The non-resampling approach is that a dichotomous variable with mean 0.03 in a sample of 112 has SD = sqrt(0.03*0.97/112) = 0.0161, so the 95%CI (assuming normality) would be 0.03 +- 1.96*0.0161 = [0.00,0.06] or exactly what I get with the bootstrapped estimate. I don't see BOOTSTRAP as an unimplemented function in PSPP so I think it's not even known by the devs to not exist (although one of the changelogs might suggest that Ben or John has adjusted the SAV file format to accommodate this). -Alan On 12/12/2018 9:32 AM, Elisa Pieri
wrote:
-- Alan D. Mead, Ph.D. President, Talent Algorithms Inc. science + technology = better workers http://www.alanmead.org "You're an interesting species. An interesting mix. You're capable of such beautiful dreams, and such horrible nightmares. You feel so lost, so cut off, so alone, only you're not. See, in all our searching, the only thing we've found that makes the emptiness bearable, is each other." -- Carl Sagan, Contact |
[Prev in Thread] | Current Thread | [Next in Thread] |