questions on LOGISTIC REGRESSION, to determine a comparison group
Locked
questions on LOGISTIC REGRESSION, to determine a comparison group
 
|
We are studying academic performance for those who went to a 1-year prep school vs. a comparison group of similar students who did not go to the prep school, then compare them in their freshmen and sophomore years.
I am picking up a study from last year and the author is unavailable (and I was the best choice to learn SPSS). I've gotten through "SPSS 101" and need a little help understanding this LOGISTIC REGRESSION. I did look at the General Syntax reference but that kind of made my eyes glaze over ... I see the general approach as to what is happening but need to understand more in depth how it works. My main question is "What's happening [below] to put students into PGROUP group_2020?" *Setup logic (lumps Direct Entry and Prep School all together) IF (feeder1=DIRECT-ENTRY AND classyear < 2021 AND classyear > 2016) prop_filter_2020_4years = 1. IF (feeder1=PREP-SCHOOL AND classyear < 2021 AND classyear > 2016) prop_filter_2020_4years = 1. VALUE LABELS prop_filter_2020_4years 1 'Direct Entry 2017-2020 and Prep School 2017-2020'. *creates PROB GROUPS for 2020. *Here it filters on the above variable prop_filter_2020_4years, then does Regression calculation *THIS IS MY MAIN QUESTION - What's happening here to put students into PGROUP group_2020? * prop_2020 has a label in the Data View "Predicted Probability", with no values * group_2020 has a label "Predicted Group", with the (translated) values DIRECT-ENTRY, PREP-SCHOOL, etc. FILTER by prop_filter_2020_4years. LOGISTIC REGRESSION VARIABLES feeder1 /METHOD=ENTER Ethnicity1 satmath satverbal gender1 hs_percent /SAVE=PRED (prop_2020) PGROUP (group_2020). FILTER off. *Later logic, giving me the Direct Entry Comparison Group for GPA's DO IF (feeder1 = DIRECT-ENTRY AND group_2020 = PREP-SCHOOL AND classyear < 2021 AND classyear > 2016). COMPUTE qpa_comp_only_2020 =1. *So here's my Comparison Group based on the above VALUE LABELS qpa_comp_only_2020 1 'Direct Entry Comparison Group 2017-2020' Thanks! ===================== To manage your subscription to SPSSX-L, send a message to [hidden email] (not to SPSSX-L), with no body text except the command. To leave the list, send the command SIGNOFF SPSSX-L For a list of commands to manage subscriptions, send the command INFO REFCARD |
Re: questions on LOGISTIC REGRESSION, to determine a comparison group
Administrator
|
Hi William. Unfortunately, the CSR manual is not very helpful here, as it
does not mention PGROUP specifically. https://www.ibm.com/support/knowledgecenter/en/SSLVMB_25.0.0/statistics_reference_project_ddita/spss/regression/syn_logistic_regression_save.html But after concocting a little example, I see that PGROUP (without an optional variable name in brackets) saves a new variable called PGR_1 with "Predicted group" as the value label. And the only possible values are 1 and 0. SPSS compares the predicted probability (PRE_1) to the cut-off value (default = 0.5) to determine whether PGR_1 = 1 or 0. To see how it works, try an example where you change the cutoff value. Here are the examples I used to figure it out. HTH. * Change path to show where you store the sample data sets. GET FILE = "C:\SPSSdata\survey_sample.sav". COMPUTE Male = sex EQ 1. COMPUTE c = 1. FORMATS Male c(F1). CROSSTABS sex by male. LOGISTIC REGRESSION VARIABLES Male /METHOD=ENTER age educ paeduc maeduc /SAVE=PRED PGROUP /PRINT=CI(95) /CRITERIA=PIN(0.05) POUT(0.10) ITERATE(20) CUT(0.5). LOGISTIC REGRESSION VARIABLES Male /METHOD=ENTER age educ paeduc maeduc /SAVE=PRED PGROUP /PRINT=CI(95) /CRITERIA=PIN(0.05) POUT(0.10) ITERATE(20) CUT(0.4). DESCRIPTIVES PRE_1 PGR_1 PRE_2 PGR_2. MEANS PRE_1 by PGR_1 / CELLS=MIN MAX. MEANS PRE_2 by PGR_2 / CELLS=MIN MAX. William Peck wrote > We are studying academic performance for those who went to a 1-year prep > school vs. a comparison group of similar students who did not go to the > prep school, then compare them in their freshmen and sophomore years. > > I am picking up a study from last year and the author is unavailable (and > I was the best choice to learn SPSS). I've gotten through "SPSS 101" and > need a little help understanding this LOGISTIC REGRESSION. I did look at > the General Syntax reference but that kind of made my eyes glaze over ... > > I see the general approach as to what is happening but need to understand > more in depth how it works. > > My main question is "What's happening [below] to put students into PGROUP > group_2020?" > > *Setup logic (lumps Direct Entry and Prep School all together) > IF (feeder1=DIRECT-ENTRY AND classyear < 2021 AND classyear > 2016) > prop_filter_2020_4years = 1. > IF (feeder1=PREP-SCHOOL AND classyear < 2021 AND classyear > 2016) > prop_filter_2020_4years = 1. > VALUE LABELS prop_filter_2020_4years > 1 'Direct Entry 2017-2020 and Prep School 2017-2020'. > > *creates PROB GROUPS for 2020. > *Here it filters on the above variable prop_filter_2020_4years, then does > Regression calculation > > *THIS IS MY MAIN QUESTION - What's happening here to put students into > PGROUP group_2020? > > * prop_2020 has a label in the Data View "Predicted Probability", with no > values > * group_2020 has a label "Predicted Group", with the (translated) values > DIRECT-ENTRY, PREP-SCHOOL, etc. > FILTER by prop_filter_2020_4years. > > LOGISTIC REGRESSION VARIABLES feeder1 > /METHOD=ENTER Ethnicity1 satmath satverbal gender1 hs_percent > /SAVE=PRED (prop_2020) PGROUP (group_2020). > > FILTER off. > > *Later logic, giving me the Direct Entry Comparison Group for GPA's > DO IF (feeder1 = DIRECT-ENTRY AND group_2020 = PREP-SCHOOL AND classyear < > 2021 AND classyear > 2016). > COMPUTE qpa_comp_only_2020 =1. > > *So here's my Comparison Group based on the above > VALUE LABELS qpa_comp_only_2020 > 1 'Direct Entry Comparison Group 2017-2020' > > Thanks! > > ===================== > To manage your subscription to SPSSX-L, send a message to > LISTSERV@.UGA > (not to SPSSX-L), with no body text except the > command. To leave the list, send the command > SIGNOFF SPSSX-L > For a list of commands to manage subscriptions, send the command > INFO REFCARD ----- -- Bruce Weaver [hidden email] http://sites.google.com/a/lakeheadu.ca/bweaver/ "When all else fails, RTFM." NOTE: My Hotmail account is not monitored regularly. To send me an e-mail, please use the address shown above. -- Sent from: http://spssx-discussion.1045642.n5.nabble.com/ ===================== To manage your subscription to SPSSX-L, send a message to [hidden email] (not to SPSSX-L), with no body text except the command. To leave the list, send the command SIGNOFF SPSSX-L For a list of commands to manage subscriptions, send the command INFO REFCARD
--
Bruce Weaver bweaver@lakeheadu.ca http://sites.google.com/a/lakeheadu.ca/bweaver/ "When all else fails, RTFM." PLEASE NOTE THE FOLLOWING: 1. My Hotmail account is not monitored regularly. To send me an e-mail, please use the address shown above. 2. The SPSSX Discussion forum on Nabble is no longer linked to the SPSSX-L listserv administered by UGA (https://listserv.uga.edu/). |
Re: questions on LOGISTIC REGRESSION, to determine a comparison group
|
|
Great! Thank you, I will review in detail and post questions as needed, but that seems to answer my question, i.e., "I see that PGROUP (without an optional variable name in brackets) saves a new variable called PGR_1 with "Predicted group" as the value label. And the only possible values are 1 and 0. SPSS compares the predicted probability (PRE_1) to the cut-off value (default = 0.5) to determine whether PGR_1 = 1 or 0." So I will dig into that and your example. On Mon, Feb 4, 2019 at 11:36 AM Bruce Weaver <[hidden email]> wrote: Hi William. Unfortunately, the CSR manual is not very helpful here, as it Bill Peck Information Technology Services Division Institutional Research / Business Intelligence United States Naval Academy 410-293-1475 (w) 410-207-3350 (c) **** Data is your friend ===> at home, at work, and on the battlefield **** **** Data is the fuel of your organization **** |
|
|
In reply to this post by Bruce Weaver
PGROUP and all the other variables that can be saved are listed under the CASESWISE subcommand. On Mon, Feb 4, 2019 at 9:36 AM Bruce Weaver <[hidden email]> wrote: Hi William. Unfortunately, the CSR manual is not very helpful here, as it |
Re: questions on LOGISTIC REGRESSION, to determine a comparison group
|
Administrator
|
Thanks Jon, that helps. It would be nice to have a pointer to CASEWISE in
the documentation for SAVE. https://www.ibm.com/support/knowledgecenter/en/SSLVMB_25.0.0/statistics_reference_project_ddita/spss/regression/syn_logistic_regression_casewise.html Jon Peck wrote > PGROUP and all the other variables that can be saved are listed under the > CASESWISE subcommand. > > On Mon, Feb 4, 2019 at 9:36 AM Bruce Weaver < > bruce.weaver@ > > > wrote: > >> Hi William. Unfortunately, the CSR manual is not very helpful here, as >> it >> does not mention PGROUP specifically. >> >> >> https://www.ibm.com/support/knowledgecenter/en/SSLVMB_25.0.0/statistics_reference_project_ddita/spss/regression/syn_logistic_regression_save.html > > --- snip --- ----- -- Bruce Weaver [hidden email] http://sites.google.com/a/lakeheadu.ca/bweaver/ "When all else fails, RTFM." NOTE: My Hotmail account is not monitored regularly. To send me an e-mail, please use the address shown above. -- Sent from: http://spssx-discussion.1045642.n5.nabble.com/ ===================== To manage your subscription to SPSSX-L, send a message to [hidden email] (not to SPSSX-L), with no body text except the command. To leave the list, send the command SIGNOFF SPSSX-L For a list of commands to manage subscriptions, send the command INFO REFCARD
--
Bruce Weaver bweaver@lakeheadu.ca http://sites.google.com/a/lakeheadu.ca/bweaver/ "When all else fails, RTFM." PLEASE NOTE THE FOLLOWING: 1. My Hotmail account is not monitored regularly. To send me an e-mail, please use the address shown above. 2. The SPSSX Discussion forum on Nabble is no longer linked to the SPSSX-L listserv administered by UGA (https://listserv.uga.edu/). |
|
|
True. I passed that along to the keepers of the CSR. On Mon, Feb 4, 2019 at 11:31 AM Bruce Weaver <[hidden email]> wrote: Thanks Jon, that helps. It would be nice to have a pointer to CASEWISE in |
Re: questions on LOGISTIC REGRESSION, to determine a comparison group
|
|
In reply to this post by Bruce Weaver
Worked through the example and got output, so that helps a lot. I still need to study in detail but this is what I was looking for. So is this accurate from a layman's perspective - "the comparison group is statistically within 95% of the PREP-SCHOOL class based on the factors a, b, c, d, e." Or how can it be best explained. I need to work through my own study and apply close inspection to the output and compare to your example. Thanks! On Mon, Feb 4, 2019 at 1:31 PM Bruce Weaver <[hidden email]> wrote: Thanks Jon, that helps. It would be nice to have a pointer to CASEWISE in Bill Peck Information Technology Services Division Institutional Research / Business Intelligence United States Naval Academy 410-293-1475 (w) 410-207-3350 (c) **** Data is your friend ===> at home, at work, and on the battlefield **** **** Data is the fuel of your organization **** |
Re: questions on LOGISTIC REGRESSION, to determine a comparison group
|
|
In reply to this post by William Peck
I have concerns about the following.
DIRECT-ENTRY AND PREP-SCHOOL are not valid variable names if - is a regular minus sign. Please double check your results. "*Setup logic (lumps Direct Entry and Prep School all together) IF (feeder1=DIRECT-ENTRY AND classyear < 2021 AND classyear > 2016) prop_filter_2020_4years = 1. IF (feeder1=PREP-SCHOOL AND classyear < 2021 AND classyear > 2016) prop_filter_2020_4years = 1. VALUE LABELS prop_filter_2020_4years 1 'Direct Entry 2017-2020 and Prep School 2017-2020'." ===================== To manage your subscription to SPSSX-L, send a message to [hidden email] (not to SPSSX-L), with no body text except the command. To leave the list, send the command SIGNOFF SPSSX-L For a list of commands to manage subscriptions, send the command INFO REFCARD |
Re: questions on LOGISTIC REGRESSION, to determine a comparison group
|
|
In reply to this post by William Peck
> So is this accurate from a layman's perspective - "the comparison group is statistically within 95% of the PREP-SCHOOL class based on the factors a, b, c, d, e." Or how can it be best explained.
No - This statement makes no sense to a statistician, so I can't
figure how it could make any sense to a layman who knows anything
about statistics.
Are you just trying to say that, for the test performed,
the groups do not differ by what would be the usual 5%-alpha test?
If you are trying to explain "no difference" - a good explanation would
also refer to the effect size and the effect of N. For instance, "An effect of
the size observed would become significant if the N had been < X%> larger" -
for the (multivariate) Logistic regression." That should be backed up by a
comment on the influence of testing 5 variables at a time - that is, the
"multi-test" problem. Perhaps - Further, "none of the univariate tests were
much better, with the strongest measured relation being ...."
--
Rich Ulrich
From: SPSSX(r) Discussion <[hidden email]> on behalf of Bill Peck <[hidden email]>
Sent: Monday, February 4, 2019 2:38 PM To: [hidden email] Subject: Re: questions on LOGISTIC REGRESSION, to determine a comparison group Worked through the example and got output, so that helps a lot. I still need to study in detail but this is what I was looking for.
So is this accurate from a layman's perspective - "the comparison group is statistically within 95% of the PREP-SCHOOL class based on the factors a, b, c, d, e." Or how can it be best explained.
I need to work through my own study and apply close inspection to the output and compare to your example.
Thanks!
On Mon, Feb 4, 2019 at 1:31 PM Bruce Weaver <[hidden email]> wrote:
Thanks Jon, that helps. It would be nice to have a pointer to CASEWISE in Bill Peck Information Technology Services Division Institutional Research / Business Intelligence United States Naval Academy 410-293-1475 (w) 410-207-3350 (c)
**** Data is your friend ===> at home, at work, and on the battlefield ****
**** Data is the fuel of your organization **** |
Re: questions on LOGISTIC REGRESSION, to determine a comparison group
|
Administrator
|
In reply to this post by William Peck
Bill, I might be missing something that is patently obvious to others, but
why are you messing around with LOGISTIC REGRESSION here at all? Here's how you described the main research question: Bill Peck wrote > We are studying academic performance for those who went to a 1-year prep > school vs. a comparison group of similar students who did not go to the > prep school, then compare them in their freshmen and sophomore years. Do you know which students went to a 1-year prep school and which ones did not? Do you have their grades in freshman and sophomore years? If YES to both of those questions, what is the purpose of using LOGISTIC REGRESSION? It seems that a mixed design ANOVA (or MIXED model) would suffice (perhaps controlling for other covariates). When I first skimmed your post, I wondered if you were trying to generate a propensity score that you wished to control for. Is that what you're doing? Thanks for clarifying. Bruce Bill Peck wrote > We are studying academic performance for those who went to a 1-year prep > school vs. a comparison group of similar students who did not go to the > prep school, then compare them in their freshmen and sophomore years. > > I am picking up a study from last year and the author is unavailable (and > I was the best choice to learn SPSS). I've gotten through "SPSS 101" and > need a little help understanding this LOGISTIC REGRESSION. I did look at > the General Syntax reference but that kind of made my eyes glaze over ... > > I see the general approach as to what is happening but need to understand > more in depth how it works. > > My main question is "What's happening [below] to put students into PGROUP > group_2020?" > > *Setup logic (lumps Direct Entry and Prep School all together) > IF (feeder1=DIRECT-ENTRY AND classyear < 2021 AND classyear > 2016) > prop_filter_2020_4years = 1. > IF (feeder1=PREP-SCHOOL AND classyear < 2021 AND classyear > 2016) > prop_filter_2020_4years = 1. > VALUE LABELS prop_filter_2020_4years > 1 'Direct Entry 2017-2020 and Prep School 2017-2020'. > > *creates PROB GROUPS for 2020. > *Here it filters on the above variable prop_filter_2020_4years, then does > Regression calculation > > *THIS IS MY MAIN QUESTION - What's happening here to put students into > PGROUP group_2020? > > * prop_2020 has a label in the Data View "Predicted Probability", with no > values > * group_2020 has a label "Predicted Group", with the (translated) values > DIRECT-ENTRY, PREP-SCHOOL, etc. > FILTER by prop_filter_2020_4years. > > LOGISTIC REGRESSION VARIABLES feeder1 > /METHOD=ENTER Ethnicity1 satmath satverbal gender1 hs_percent > /SAVE=PRED (prop_2020) PGROUP (group_2020). > > FILTER off. > > *Later logic, giving me the Direct Entry Comparison Group for GPA's > DO IF (feeder1 = DIRECT-ENTRY AND group_2020 = PREP-SCHOOL AND classyear < > 2021 AND classyear > 2016). > COMPUTE qpa_comp_only_2020 =1. > > *So here's my Comparison Group based on the above > VALUE LABELS qpa_comp_only_2020 > 1 'Direct Entry Comparison Group 2017-2020' > > Thanks! > > ===================== > To manage your subscription to SPSSX-L, send a message to > LISTSERV@.UGA > (not to SPSSX-L), with no body text except the > command. To leave the list, send the command > SIGNOFF SPSSX-L > For a list of commands to manage subscriptions, send the command > INFO REFCARD ----- -- Bruce Weaver [hidden email] http://sites.google.com/a/lakeheadu.ca/bweaver/ "When all else fails, RTFM." NOTE: My Hotmail account is not monitored regularly. To send me an e-mail, please use the address shown above. -- Sent from: http://spssx-discussion.1045642.n5.nabble.com/ ===================== To manage your subscription to SPSSX-L, send a message to [hidden email] (not to SPSSX-L), with no body text except the command. To leave the list, send the command SIGNOFF SPSSX-L For a list of commands to manage subscriptions, send the command INFO REFCARD
--
Bruce Weaver bweaver@lakeheadu.ca http://sites.google.com/a/lakeheadu.ca/bweaver/ "When all else fails, RTFM." PLEASE NOTE THE FOLLOWING: 1. My Hotmail account is not monitored regularly. To send me an e-mail, please use the address shown above. 2. The SPSSX Discussion forum on Nabble is no longer linked to the SPSSX-L listserv administered by UGA (https://listserv.uga.edu/). |
Re: questions on LOGISTIC REGRESSION, to determine a comparison group
|
|
In reply to this post by Rich Ulrich
|
«
Return to SPSSX Discussion
|
1 view|%1 views
| Free forum by Nabble | Edit this page |