Re: ensuring confidentiality (fwd)
Thank you all for your ideas. I don't know whether I have a solution to my problem yet, but your input has certainly been helpful in getting me pointed in the right direction. Thanks again. David Schmalz = Instructions for joining and leaving this list and remarks about the problem of INAPPROPRIATE MESSAGES are available at http://jse.stat.ncsu.edu/ =
Re: ensuring confidentiality (fwd)
On Fri, 4 Aug 2000, dennis roberts wrote: At 09:59 AM 8/4/00 -0400, Bob Hayden wrote, replying to Peter Lewycky: I've yet to meet an (adult) respondent who did not know his mother's maiden name and her birthdate. :) We must meet sometime!-) can i come too? i remember her maiden name but not her birth date in fact, this is an interesting question ... take a random sample of adults ... say age 40 to 50 ... Why so restricted a range of ages? See below... and ask these two questions ... wonder which answer has the higher p value of being correct? i predict the name does ... Quite likely. But add to your survey questions at least one dealing with the celebration of birthdays in the family. I conjecture that offspring of families where birthdays are routinely celebrated (for the adults as well as for the children) will more often -- nearly always, actually -- recall parental birthdays. Interesting to compare father's and mother's birthdays, as well... Your age restriction would eliminate me, thus supplying a small bias in favor of your expectation (since I happen to know both parents' birthdays, and mother's (and both grandmothers') maiden names). Perhaps your survey should also inquire how long the FAMILY has resided in the present town, state, and nation ... Donald F. Burrill [EMAIL PROTECTED] 348 Hyde Hall, Plymouth State College, [EMAIL PROTECTED] MSC #29, Plymouth, NH 03264 603-535-2597 184 Nashua Road, Bedford, NH 03110 603-471-7128 = Instructions for joining and leaving this list and remarks about the problem of INAPPROPRIATE MESSAGES are available at http://jse.stat.ncsu.edu/ =
Re: ensuring confidentiality (fwd)
- Forwarded message from Peter Lewycky - I've yet to meet an (adult) respondent who did not know his mother's maiden name and her birthdate. :) - End of forwarded message from Peter Lewycky - We must meet sometime!-) _ | |Robert W. Hayden | | Work: Department of Mathematics / |Plymouth State College MSC#29 | |Plymouth, New Hampshire 03264 USA | * |fax (603) 535-2943 /| Home: 82 River Street (use this in the summer) | )Ashland, NH 03217 L_/(603) 968-9914 (use this year-round) Map of New[EMAIL PROTECTED] (works year-round) Hampshire http://mathpc04.plymouth.edu (works year-round) = Instructions for joining and leaving this list and remarks about the problem of INAPPROPRIATE MESSAGES are available at http://jse.stat.ncsu.edu/ =
Re: Ensuring confidentiality
On 3 Aug 2000, DavidS9307 wrote: I would like to collect data in a school on a survey form where the respondents enter only a code number to identify themselves. I would like the code number to be something that the participants will be able to remember for follow-up data collection in the near future, but I would also like to protect against identifying the participant even if the forms and coding scheme were disclosed, so I don't want to use birthdates, names of relatives, or SSNs. Surely you also want the codes to be unique, so that your follow-up analyses do not confuse Maxine with Raymond simply because they happened to produce the same code. I had been thinking of using some number not generally known to anyone other than the participant, such as the last digit of their locker number at the school, "Last digit"? You ARE planning to have a sample size larger than ten, aren't you? to disguise a universal identifier by adding it to each of the last four digits of the SSN, but that seems rather clumsy. I take it that "adding" refers to doing arithmetic, not to merely appending a common value (which would not disguise things much!). But there is still the question of uniqueness: can you be guaranteed that there are no duplicates of the last 4 digits? I doubt it. Does anyone have a suggestion for a good coding scheme? Thanks to all for any recommendations. Generate a set of, say, 3- or 4-digit numbers. Print each on one side of a card. Shuffle the cards. Take the cards (or a sufficiently large subset of them) into the classroom where you introduce the survey to the participants. Visibly shuffle them again, and deal them (face down), or have assistants deal them, one to a customer. Explain why you need the numbers, ask them to memorize their own code (and/or squirrel the card away in a safe place where they can find it later), and write it in the space provided for the purpose on the survey response forms. If you are dealing with classroom-sized groups of participants (i.e., fewer than 52 persons per subgroup) you could use ordinary playing cards plus a precoded identifier for the subgroup. Makes a 2-character identifier, possibly easier to remember than an arbitrary 3- or 4-digit number (though people seem to do all right with PINs for their ATM cards): AH for ace of hearts, 3S for trey of spades, etc. And you'd have to ask them to remember the subgroup identifier as well, but perhaps you could use a single alphabetic character for that (or a number if there are ten or fewer groups). (Make sure you remove the jokers etc. before you shuffle deal!) Donald F. Burrill [EMAIL PROTECTED] 348 Hyde Hall, Plymouth State College, [EMAIL PROTECTED] MSC #29, Plymouth, NH 03264 603-535-2597 184 Nashua Road, Bedford, NH 03110 603-471-7128 = Instructions for joining and leaving this list and remarks about the problem of INAPPROPRIATE MESSAGES are available at http://jse.stat.ncsu.edu/ =
Re: Ensuring confidentiality
I've used mother's maiden name (abbreviated if too long or padded if too short)+ birthdate of mother or respondent. Donald Burrill wrote: On 3 Aug 2000, DavidS9307 wrote: I would like to collect data in a school on a survey form where the respondents enter only a code number to identify themselves. I would like the code number to be something that the participants will be able to remember for follow-up data collection in the near future, but I would also like to protect against identifying the participant even if the forms and coding scheme were disclosed, so I don't want to use birthdates, names of relatives, or SSNs. Surely you also want the codes to be unique, so that your follow-up analyses do not confuse Maxine with Raymond simply because they happened to produce the same code. I had been thinking of using some number not generally known to anyone other than the participant, such as the last digit of their locker number at the school, "Last digit"? You ARE planning to have a sample size larger than ten, aren't you? to disguise a universal identifier by adding it to each of the last four digits of the SSN, but that seems rather clumsy. I take it that "adding" refers to doing arithmetic, not to merely appending a common value (which would not disguise things much!). But there is still the question of uniqueness: can you be guaranteed that there are no duplicates of the last 4 digits? I doubt it. Does anyone have a suggestion for a good coding scheme? Thanks to all for any recommendations. Generate a set of, say, 3- or 4-digit numbers. Print each on one side of a card. Shuffle the cards. Take the cards (or a sufficiently large subset of them) into the classroom where you introduce the survey to the participants. Visibly shuffle them again, and deal them (face down), or have assistants deal them, one to a customer. Explain why you need the numbers, ask them to memorize their own code (and/or squirrel the card away in a safe place where they can find it later), and write it in the space provided for the purpose on the survey response forms. If you are dealing with classroom-sized groups of participants (i.e., fewer than 52 persons per subgroup) you could use ordinary playing cards plus a precoded identifier for the subgroup. Makes a 2-character identifier, possibly easier to remember than an arbitrary 3- or 4-digit number (though people seem to do all right with PINs for their ATM cards): AH for ace of hearts, 3S for trey of spades, etc. And you'd have to ask them to remember the subgroup identifier as well, but perhaps you could use a single alphabetic character for that (or a number if there are ten or fewer groups). (Make sure you remove the jokers etc. before you shuffle deal!) Donald F. Burrill [EMAIL PROTECTED] 348 Hyde Hall, Plymouth State College, [EMAIL PROTECTED] MSC #29, Plymouth, NH 03264 603-535-2597 184 Nashua Road, Bedford, NH 03110 603-471-7128 = Instructions for joining and leaving this list and remarks about the problem of INAPPROPRIATE MESSAGES are available at http://jse.stat.ncsu.edu/ = = Instructions for joining and leaving this list and remarks about the problem of INAPPROPRIATE MESSAGES are available at http://jse.stat.ncsu.edu/ =
Re: Ensuring confidentiality
On 03 Aug 2000 06:45:45 GMT, [EMAIL PROTECTED] (DavidS9307) wrote: I would like to collect data in a school on a survey form where the respondents enter only a code number to identify themselves. I would like the code number to be something that the participants will be able to remember for follow-up data collection in the near future, but I would also like to protect against identifying the participant even if the forms and coding scheme were disclosed, so I don't want to use birthdates, names of relatives, or SSNs. I had been thinking of using some number not generally known to anyone other than the participant, such as the last digit of their locker number at the school, to disguise a universal identifier by adding it to each of the last four digits of the SSN, but that seems rather clumsy. Does anyone have a suggestion for a good coding scheme? Thanks to all for any recommendations. I don't know if you can fit it to your scheme, but one scheme that I heard of - I have never seen it done - asked people to write down the serial number from a dollar bill. Maybe you could hand out the bills? Respondents can write down their numbers, or save their bills. -- Rich Ulrich, [EMAIL PROTECTED] http://www.pitt.edu/~wpilib/index.html = Instructions for joining and leaving this list and remarks about the problem of INAPPROPRIATE MESSAGES are available at http://jse.stat.ncsu.edu/ =
Re: ensuring confidentiality
The original question was how to come up with a unique identification that would be known only to the respondent and would be remembered. As long as they put in 4 Latin characters followed by 6 Arabic numbers (that could pass for a date ... I don't check). Maybe they used their girl friend's name and gave the date of Roosevelt's election. I don't care than they remember what they did and could reconstruct the id. I've yet to meet an (adult) respondent who did not know his mother's maiden name and her birthdate. :) Bob Hayden wrote: You must have a different audience than I if you think they can spell other people's names, remember numbers, identify and encode playing cards, know their mother's maiden name and birthdate, etc. The task needs to be one they are capable of and motivated to attempt. Maybe the survey could include a copy of (part of) a dollar bill. The serial number on the bill is the ID. Respondents are asked to copy it down and put it on each subsequent response. When the study is over, everyone who has submitted a full set of responses gets the real dollar bill. It might need to be $5, $10 or $20 or even higher depending on the age and wealth of the respondent, and how much work is involved in answering all the questions. _ | |Robert W. Hayden | | Work: Department of Mathematics / |Plymouth State College MSC#29 | |Plymouth, New Hampshire 03264 USA | * |fax (603) 535-2943 /| Home: 82 River Street (use this in the summer) | )Ashland, NH 03217 L_/(603) 968-9914 (use this year-round) Map of New[EMAIL PROTECTED] (works year-round) Hampshire http://mathpc04.plymouth.edu (works year-round) = Instructions for joining and leaving this list and remarks about the problem of INAPPROPRIATE MESSAGES are available at http://jse.stat.ncsu.edu/ = = Instructions for joining and leaving this list and remarks about the problem of INAPPROPRIATE MESSAGES are available at http://jse.stat.ncsu.edu/ =
