About elbarto

elbarto · ‎10-14-2023

I have data like the following (but millions of rows, this is just an excerpt): DATA have; input individual & $char10. firm year num; length individual $10; DATALINES; 100004-77 1001 2004 3 100004-77 1004 2004 3 100004-77 1006 2004 3 100004-77 1008 2004 3 100004-77 1013 2004 3 100004-77 1015 2004 3 119561-14 1003 2006 5 119561-14 1004 2006 5 119561-14 1005 2006 5 119561-14 1008 2006 5 119561-14 1009 2006 5 119561-14 1012 2006 5 119561-14 1014 2006 5 119561-14 1015 2006 5 ; RUN; I want to randomly select a sample for each individual year group based on the number in num, e.g., for individual=100004-77 and year=2004, I want to pick a random sample of size 3. I tried the following, but it does not work: proc surveyselect data=have out=want seed=1234 method=srs n=num; strata individual year; run; I need to automate this because I have considerably many more rows, so I can't simply create two different datasets based on each individual-year combination then manually enter the sample size. Thank you.

elbarto · ‎10-13-2023

I have data that is structured in a matrix format exactly like this (except with more rows and columns). I've copied the raw data as I am not even sure how to load this into SAS. firm 1001 1001 1001 1001 1001 1001 1001 1001 1001 1001 1003 1003 1003 1003 1003 1003 1003 1003 1003 year 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2010 2011 2012 2013 2014 2015 2016 2017 2018 individual event_year 103307-86 2010 0.3 0.0 1.0 0.6 0.0 0.0 0.9 0.8 0.8 0.0 1.0 0.7 0.1 0.1 0.7 0.4 0.3 0.6 0.3 42185-26 2005 0.8 0.6 0.3 0.7 0.8 0.4 0.7 0.2 0.1 0.6 0.2 0.4 0.8 0.1 0.4 0.5 1.0 0.4 0.5 42185-26 2009 0.6 0.7 0.9 0.4 0.5 0.1 0.9 0.7 0.9 0.6 0.6 0.7 0.3 0.5 0.9 0.1 1.0 0.7 0.7 Across the columns are two headers, firm identifiers (e.g., 1001, 1003) and year identifiers for each firm (e.g., 2005 to 2014 for firm=1001 and 2010 to 2018 for firm=1003). Across the rows, I have identifiers for individuals and an event that occurs in a given year for that individual ("event_year"). What I'd like to do is transpose the data into the following format: DATA want; input individual & $char10. event_year firm year score; length individual $10; DATALINES; 103307-86 2010 1001 2005 0.3 103307-86 2010 1001 2006 0.0 103307-86 2010 1001 2007 1.0 103307-86 2010 1001 2008 0.6 103307-86 2010 1001 2009 0.0 103307-86 2010 1001 2010 0.0 103307-86 2010 1001 2011 0.9 103307-86 2010 1001 2012 0.8 103307-86 2010 1001 2013 0.8 103307-86 2010 1001 2014 0.0 103307-86 2010 1003 2010 1.0 103307-86 2010 1003 2011 0.7 103307-86 2010 1003 2012 0.1 103307-86 2010 1003 2013 0.1 103307-86 2010 1003 2014 0.7 103307-86 2010 1003 2015 0.4 103307-86 2010 1003 2016 0.3 103307-86 2010 1003 2017 0.6 103307-86 2010 1003 2018 0.3 42185-26 2005 1001 2005 0.8 42185-26 2005 1001 2006 0.6 42185-26 2005 1001 2007 0.3 42185-26 2005 1001 2008 0.7 42185-26 2005 1001 2009 0.8 42185-26 2005 1001 2010 0.4 42185-26 2005 1001 2011 0.7 42185-26 2005 1001 2012 0.2 42185-26 2005 1001 2013 0.1 42185-26 2005 1001 2014 0.6 42185-26 2005 1003 2010 0.2 42185-26 2005 1003 2011 0.4 42185-26 2005 1003 2012 0.8 42185-26 2005 1003 2013 0.1 42185-26 2005 1003 2014 0.4 42185-26 2005 1003 2015 0.5 42185-26 2005 1003 2016 1.0 42185-26 2005 1003 2017 0.4 42185-26 2005 1003 2018 0.5 42185-26 2009 1001 2005 0.6 42185-26 2009 1001 2006 0.7 42185-26 2009 1001 2007 0.9 42185-26 2009 1001 2008 0.4 42185-26 2009 1001 2009 0.5 42185-26 2009 1001 2010 0.1 42185-26 2009 1001 2011 0.9 42185-26 2009 1001 2012 0.7 42185-26 2009 1001 2013 0.9 42185-26 2009 1001 2014 0.6 42185-26 2009 1003 2010 0.6 42185-26 2009 1003 2011 0.7 42185-26 2009 1003 2012 0.3 42185-26 2009 1003 2013 0.5 42185-26 2009 1003 2014 0.9 42185-26 2009 1003 2015 0.1 42185-26 2009 1003 2016 1.0 42185-26 2009 1003 2017 0.7 42185-26 2009 1003 2018 0.7 ; RUN; As you can see, I want to transpose the data into "vertical" format, where each row corresponds to a individual, event_year, firm, and year observation. I've created a variable called "score" that contains the value in the matrix. Thank you.

elbarto · ‎10-12-2023

Thank you! I managed to tweak your code to get it 🙂 data temp; set have(keep=individual firm individual_exit_year); run; proc sort data=temp out=base nodup; by individual firm individual_exit_year; run; proc sort data=have out=temp2(keep=individual investment_year) nodupkey; by individual investment_year; run; data base2; set base; do year=2005 to 2023; output; end; run; proc sql; create table base3 as select distinct a.*, b.investment_year, c.investment_year as inv_year from base2 a left join have b on a.individual=b.individual and a.firm=b.firm and a.individual_exit_year=b.individual_exit_year and a.year=b.investment_year left join temp2 c on a.individual=c.individual and a.year=c.investment_year order by firm, individual, year ; quit; data base3; set base3; if investment_year^=. then do; if investment_year = inv_year then treat=1; end; if inv_year^=. and investment_year=. then treat=0; run; data base3; set base3; by firm individual; if first.individual then treat_fill=treat; else treat_fill=coalesce(treat,treat_fill); /** Note order matters treat, treat_fill **/ retain treat_fill; run; data base3; set base3; if treat_fill=. then treat_fill=0; if individual_exit_year^=. then do; if year>=individual_exit_year then treat_fill=0; end; run;

elbarto · ‎10-11-2023

I have a dataset which looks like the following: DATA have; input individual investment_year firm individual_exit_year; DATALINES; 1 2006 1004 . 1 2006 1005 . 1 2006 1006 . 1 2007 1004 . 1 2007 1005 . 1 2007 1007 . 1 2010 1003 . 1 2010 1006 . 1 2010 1007 . 1 2015 1004 . 1 2015 1005 . 1 2015 1007 . 2 2005 1003 2021 2 2005 1006 2021 2 2005 1008 2021 2 2008 1004 2021 2 2008 1005 2021 2 2008 1006 2021 2 2013 1005 2021 2 2013 1006 2021 2 2013 1008 2021 2 2019 1003 2021 2 2019 1004 2021 2 2019 1005 2021 3 2010 1003 2010 ; RUN; I want to create the following dataset based on the above: DATA want; input firm year individual treatment; DATALINES; 1003 2005 1 0 1003 2006 1 0 1003 2007 1 0 1003 2008 1 0 1003 2009 1 0 1003 2010 1 1 1003 2011 1 1 1003 2012 1 1 1003 2013 1 1 1003 2014 1 1 1003 2015 1 0 1003 2016 1 0 1003 2017 1 0 1003 2018 1 0 1003 2019 1 0 1003 2020 1 0 1003 2021 1 0 1003 2022 1 0 1003 2023 1 0 1004 2005 1 0 1004 2006 1 1 1004 2007 1 1 1004 2008 1 1 1004 2009 1 1 1004 2010 1 0 1004 2011 1 0 1004 2012 1 0 1004 2013 1 0 1004 2014 1 0 1004 2015 1 1 1004 2016 1 1 1004 2017 1 1 1004 2018 1 1 1004 2019 1 1 1004 2020 1 1 1004 2021 1 1 1004 2022 1 1 1004 2023 1 1 1005 2005 1 0 1005 2006 1 1 1005 2007 1 1 1005 2008 1 1 1005 2009 1 1 1005 2010 1 0 1005 2011 1 0 1005 2012 1 0 1005 2013 1 0 1005 2014 1 0 1005 2015 1 1 1005 2016 1 1 1005 2017 1 1 1005 2018 1 1 1005 2019 1 1 1005 2020 1 1 1005 2021 1 1 1005 2022 1 1 1005 2023 1 1 1006 2005 1 0 1006 2006 1 1 1006 2007 1 0 1006 2008 1 0 1006 2009 1 0 1006 2010 1 1 1006 2011 1 1 1006 2012 1 1 1006 2013 1 1 1006 2014 1 1 1006 2015 1 0 1006 2016 1 0 1006 2017 1 0 1006 2018 1 0 1006 2019 1 0 1006 2020 1 0 1006 2021 1 0 1006 2022 1 0 1006 2023 1 0 1007 2005 1 0 1007 2006 1 0 1007 2007 1 1 1007 2008 1 1 1007 2009 1 1 1007 2010 1 1 1007 2011 1 1 1007 2012 1 1 1007 2013 1 1 1007 2014 1 1 1007 2015 1 1 1007 2016 1 1 1007 2017 1 1 1007 2018 1 1 1007 2019 1 1 1007 2020 1 1 1007 2021 1 1 1007 2022 1 1 1007 2023 1 1 1003 2005 2 1 1003 2006 2 1 1003 2007 2 1 1003 2008 2 0 1003 2009 2 0 1003 2010 2 0 1003 2011 2 0 1003 2012 2 0 1003 2013 2 0 1003 2014 2 0 1003 2015 2 0 1003 2016 2 0 1003 2017 2 0 1003 2018 2 0 1003 2019 2 1 1003 2020 2 1 1003 2021 2 0 1003 2022 2 0 1003 2023 2 0 1004 2005 2 0 1004 2006 2 0 1004 2007 2 0 1004 2008 2 1 1004 2009 2 1 1004 2010 2 1 1004 2011 2 1 1004 2012 2 1 1004 2013 2 0 1004 2014 2 0 1004 2015 2 0 1004 2016 2 0 1004 2017 2 0 1004 2018 2 0 1004 2019 2 1 1004 2020 2 1 1004 2021 2 0 1004 2022 2 0 1004 2023 2 0 1005 2005 2 0 1005 2006 2 0 1005 2007 2 0 1005 2008 2 1 1005 2009 2 1 1005 2010 2 1 1005 2011 2 1 1005 2012 2 1 1005 2013 2 1 1005 2014 2 1 1005 2015 2 1 1005 2016 2 1 1005 2017 2 1 1005 2018 2 1 1005 2019 2 1 1005 2020 2 1 1005 2021 2 0 1005 2022 2 0 1005 2023 2 0 1006 2005 2 1 1006 2006 2 1 1006 2007 2 1 1006 2008 2 1 1006 2009 2 1 1006 2010 2 1 1006 2011 2 1 1006 2012 2 1 1006 2013 2 1 1006 2014 2 1 1006 2015 2 1 1006 2016 2 1 1006 2017 2 1 1006 2018 2 1 1006 2019 2 0 1006 2020 2 0 1006 2021 2 0 1006 2022 2 0 1006 2023 2 0 1008 2005 2 1 1008 2006 2 1 1008 2007 2 1 1008 2008 2 0 1008 2009 2 0 1008 2010 2 0 1008 2011 2 0 1008 2012 2 0 1008 2013 2 1 1008 2014 2 1 1008 2015 2 1 1008 2016 2 1 1008 2017 2 1 1008 2018 2 1 1008 2019 2 0 1008 2020 2 0 1008 2021 2 0 1008 2022 2 0 1008 2023 2 0 1003 2005 3 0 1003 2006 3 0 1003 2007 3 0 1003 2008 3 0 1003 2009 3 0 1003 2010 3 0 1003 2011 3 0 1003 2012 3 0 1003 2013 3 0 1003 2014 3 0 1003 2015 3 0 1003 2016 3 0 1003 2017 3 0 1003 2018 3 0 1003 2019 3 0 1003 2020 3 0 1003 2021 3 0 1003 2022 3 0 1003 2023 3 0 ; RUN; I want to create a timeseries of a new variable called "treatment" for each firm from 2005 to 2023 based on the investment_year for each individual in the "have" dataset. The rules to create this "treatment" variable is quite simple, but best illustrated using an example. Consider firm=1004 and individual=1 in "have". This individual invests in this firm in 2006 (i.e., in investment_year=2006, we see firm=1004 appears for individual=1), the individual invests in this firm again in 2007 (i.e., firm=1004 appears again for individual 1 when investment_year=2007), does NOT invest in it in 2010 (i.e., firm=1004 does NOT appear for individual 1 in investment_year=2010), and then invests in it again in 2015 (i.e., firm=1004 appears again for individual 1 when investment_year=2015). The "treatment" variable in "want"is a binary variable that takes either the value of 0 or 1 based on when the individual invests in the given firm. So, for firm=1004 and individual=1, I want to create a timeseries from 2005 to 2023 where "treatment" starts with a value of 0 in 2005 but takes on a value of 1 from 2006 (because individual 1 invests in this firm), it stays on as 1 until the next round of investment, which is in 2007, it will stay as 1 from 2007 to 2010, but will switch back to 0 from 2010 since the individual does not invest in this firm from 2010, but it will switch back to 1 from 2015 onwards since the individual invests in this firm again. Note that if individual_exit_year is empty, then the value of treatment stays from the last investment_year until 2023. In this case, because individual 1 invests in firm=1004 in investment_year=2015 and this individual has an empty individual_exit_year, treatment will equal to 1 from 2015 all the way until 2023. However, if individual_exit_year is NOT empty, then the value of treatment will equal to 0 from whenever the year in individual_exit_year is. As an example, consider firm=1003 and individual=2. This individual invests in this firm in 2005, so treatment will take a value of 1 from 2005 until 2008 when it will switch to 0 because the individual does NOT invest in this firm in investment_year=2008. Treatment will keep maintaining a value of 0 through to 2013 (since the individual still does NOT invest in this firm in investment_year=2013). But, treatment will take a value of 1 from 2019 (since the individual invests in this firm in 2019), but because individual_exit_year=2021 for individual 2, treatment will only take a value of 1 until 2021, where it will switch back to 0 until 2023. Note that if investment_year and individual_exit_year are the same (as is the case for individual=3), the individual_exit_year takes precedence, i.e., treatment will take a value of 0 until 2023.

elbarto · ‎09-26-2023

Thanks that works! I also tried the following, which works as well: type_nc = compress(type); CVC = 0 < indexw(type_nc,'CorporateVentureCapital',", "); VC = 0 < indexw(type_nc,'VentureCapital',", ");

elbarto · ‎09-26-2023

Thank you. However, when I try the following: data want; set have; CVC = 0 < indexw(Type,'Corporate Venture Capital'); VC= 0 < indexw(Type,'Venture Capital'); run; It does not seem to be picking up the "Corporate Venture Capital", for example, CVC is 0 for both investorID=1002 and 1004, when it should be 1.

elbarto · ‎09-25-2023

Sorry, I mean the entire string "Corporate Venture Capital" to get a match. Likewise for the entire string "Venture Capital".

elbarto · ‎09-25-2023

I have an example dataset as follows: DATA have; input InvestorID Type & $char1000.; length Type $1000; DATALINES; 1001 Accelerator/Incubator 1002 Accelerator/Incubator, Angel (individual), Angel Group, Asset Manager, Business Development Company, Corporate Development, Corporate Venture Capital, Corporation, Family Office, Fund of Funds, Fundless Sponsor, Government, Growth/Expansion, Hedge Fund, Holding Company, Impact Investing, Infrastructure, Investment Bank, Leasing, Lender/Debt Provider, Limited Partner, Merchant Banking Firm, Mezzanine, Mutual Fund, Not-For-Profit Venture Capital, Other Private Equity, PE-Backed Company, PE/Buyout, Real Estate, SBIC, Secondary Buyer, Sovereign Wealth Fund, University, VC-Backed Company, Venture Capital 1003 Accelerator/Incubator, Angel (individual), Business Development Company, Venture Capital 1004 Corporate Venture Capital, Fund of Funds ; RUN; What I want to do is produce the following dataset: DATA want; input InvestorID Type & $char1000. CVC VC; length Type $1000; DATALINES; 1001 Accelerator/Incubator 0 0 1002 Accelerator/Incubator, Angel (individual), Angel Group, Asset Manager, Business Development Company, Corporate Development, Corporate Venture Capital, Corporation, Family Office, Fund of Funds, Fundless Sponsor, Government, Growth/Expansion, Hedge Fund, Holding Company, Impact Investing, Infrastructure, Investment Bank, Leasing, Lender/Debt Provider, Limited Partner, Merchant Banking Firm, Mezzanine, Mutual Fund, Not-For-Profit Venture Capital, Other Private Equity, PE-Backed Company, PE/Buyout, Real Estate, SBIC, Secondary Buyer, Sovereign Wealth Fund, University, VC-Backed Company, Venture Capital 1 1 1003 Accelerator/Incubator, Angel (individual), Business Development Company, Venture Capital 0 1 1004 Corporate Venture Capital, Fund of Funds 1 0 ; RUN; The CVC variable takes on a value of 1 if in Type, the word Corporate Venture Capital appears, otherwise it is 0. Likewise the VC variable takes on a value of 1 if in Type, the word Venture Capital appears, otherwise it is 0. These words Corporate Venture Capital and Venture Capital are always spelt with the same capitalization.

elbarto · ‎06-17-2023

I have the following dataset: DATA have; input year person_id firm_id; DATALINES; 2000 1 405 2001 1 405 2002 1 405 2003 1 302 2004 1 302 2005 1 302 2006 1 405 2007 1 405 2008 1 405 2009 1 543 2010 1 543 2005 2 112 2006 2 112 2007 2 112 2010 2 112 2011 2 405 2012 2 543 2013 2 112 2014 2 112 2015 2 112 2016 2 112 2017 2 112 2018 2 112 2019 2 112 2001 3 405 2004 3 405 2005 3 302 2010 3 302 2011 3 405 2014 3 405 2015 3 405 2016 3 405 2000 4 302 2005 4 405 2010 4 112 ; RUN; I want to create a variable called tenure to produce the following: DATA want; input year person_id firm_id tenure; DATALINES; 2000 1 405 0 2001 1 405 1 2002 1 405 2 2003 1 302 0 2004 1 302 1 2005 1 302 2 2006 1 405 0 2007 1 405 1 2008 1 405 2 2009 1 543 0 2010 1 543 1 2005 2 112 0 2006 2 112 1 2007 2 112 2 2010 2 112 5 2011 2 405 0 2012 2 543 0 2013 2 112 0 2014 2 112 1 2015 2 112 2 2016 2 112 3 2017 2 112 4 2018 2 112 5 2019 2 112 6 2001 3 405 0 2004 3 405 3 2005 3 302 0 2010 3 302 5 2011 3 405 0 2014 3 405 3 2015 3 405 4 2016 3 405 5 2000 4 302 0 2005 4 405 0 2010 4 112 0 ; RUN; Basically, tenure is constructed as the number of consecutive years that the same firm_id has served the person_id. For example, firm_id=405 serves person_id=1 from 2000 to 2002, but then person_id=1 switches to firm_id=302 in 2003, so the tenure variable resets to 0. In 2006, the person switches back to firm_id=405, even though this firm has served this person before, tenure is defined as consecutive years of service, so tenure begins from 0 again in 2006. Note that there could be gaps in years, if there are gaps, and the firm_id did not change, then we assume the same firm_id has served the person_id. E.g., person_id=3 and firm_id=405 in 2001 and 2004. Even though there is a gap, since the firm_id does not change, tenure is 3 in 2004. If the firm_id changes when there is a gap, then tenure = 0. E.g., person_id=4.

elbarto · ‎04-27-2023

Thanks Kurt, however when I tested the code, all the fwd_sum1 and back_sum1 are missing. This is after correcting the typo in fwd_sum2 = sum(fdw_sum1,0); Any ideas why?

elbarto · ‎04-27-2023

I have data that looks like the following: DATA have; input unit year variable; DATALINES; 1 2000 0.12 1 2001 0.12 1 2002 0.98 1 2003 0.89 1 2004 0.59 1 2005 0.70 1 2006 0.17 1 2007 0.71 1 2008 0.70 1 2009 0.85 1 2010 0.86 1 2011 0.25 1 2012 0.60 1 2013 0.11 1 2014 0.39 1 2015 0.28 1 2016 0.04 1 2017 0.18 2 1978 0.97 2 1979 0.70 2 1981 0.30 2 1983 0.15 2 1984 0.23 2 1986 0.02 2 1987 0.90 2 1988 0.22 2 1989 0.70 2 1990 0.13 2 1991 0.95 2 1992 0.23 2 1994 0.57 2 1995 0.83 2 1996 0.89 2 1998 0.39 3 2010 0.33 3 2011 0.14 3 2014 0.09 ; RUN; I want to create the following dataset with four variables: fwd_sum1 fwd_sum2 back_sum1 back_sum2 as follows: DATA want; input unit year variable fwd_sum1 fwd_sum2 back_sum1 back_sum2; DATALINES; 1 2000 0.12 6.43 6.43 . 0 1 2001 0.12 5.84 5.84 . 0 1 2002 0.98 5.14 5.14 . 0 1 2003 0.89 4.97 4.97 . 0 1 2004 0.59 4.26 4.26 0.12 0.12 1 2005 0.70 3.56 3.56 0.24 0.24 1 2006 0.17 2.71 2.71 1.22 1.22 1 2007 0.71 1.85 1.85 2.11 2.11 1 2008 0.70 1.60 1.60 2.70 2.70 1 2009 0.85 1.00 1.00 3.40 3.40 1 2010 0.86 0.89 0.89 3.57 3.57 1 2011 0.25 0.50 0.50 4.28 4.28 1 2012 0.60 0.22 0.22 4.98 4.98 1 2013 0.11 0.18 0.18 5.83 5.83 1 2014 0.39 . 0 6.69 6.69 1 2015 0.28 . 0 6.94 6.94 1 2016 0.04 . 0 7.54 7.54 1 2017 0.18 . 0 7.65 7.65 2 1978 0.97 6.21 6.21 . 0 2 1979 0.70 6.21 6.21 . 0 2 1981 0.30 5.83 5.83 . 0 2 1983 0.15 5.81 5.81 1.67 1.67 2 1984 0.23 4.91 4.91 1.67 1.67 2 1986 0.02 3.99 3.99 1.97 1.97 2 1987 0.90 3.86 3.86 2.12 2.12 2 1988 0.22 2.91 2.91 2.35 2.35 2 1989 0.70 2.68 2.68 2.35 2.35 2 1990 0.13 2.68 2.68 2.37 2.37 2 1991 0.95 2.11 2.11 3.27 3.27 2 1992 0.23 1.28 1.28 3.49 3.49 2 1994 0.57 0.39 0.39 4.32 4.32 2 1995 0.83 . 0 5.27 5.27 2 1996 0.89 . 0 5.50 5.5 2 1998 0.39 . 0 6.07 6.07 3 2010 0.33 0.09 0.09 . 0 3 2011 0.14 . 0 . 0 3 2014 0.09 . 0 0.33 0.33 ; RUN; fwd_sum1 is a "forward" sum defined as follows: For each year of a given unit, fwd_sum1 is obtained by the sum of the variable values beginning from four years into the future until the end of the sample period for that unit. For example, for unit 1, year 2000, fwd_sum1 is obtained by adding all values of variable from 2004 (four years from 2000) until 2017 (end of sample period for unit 1). However, note that when year=2014, there is no 2018 for this unit, so fwd_sum1 takes on a missing value. However, note that the panel of data is not balanced, so some years may be missing for some units, for these missing years, the value of variable is assumed to be missing. For example, for unit 2, year=1978. Four years into the future is 1982, but this year is missing for this unit, so the 1982 value of variable is missing. fwd_sum2 is defined almost the same as fwd_sum1 but if the entire sum given by fwd_sum1 is missing, then set the value to 0. back_sum1 is a "backward" sum defined as follows: For each year of a given unit, back_sum1 is obtained by the sum of the variable values beginning from four years into the past until the start of the sample period for that unit. For example, for unit=1, year=2017, back_sum1 is obtained by adding all values of variable from 2013 (four years in the past) until 2000 (start of sample period for unit 1). Again, note that when year=2003, there is no 1999 for this unit, so back_sum1 is just missing. back_sum2 is defined almost the same as back_sum1 except if the entire sum given by back_sum1 is missing, then set the value to 0. Finally, the above example is given for four years in the future (for the fwd sums) and four years in the past (for the back sums). Ideally, I would like a more general result, i.e., N years in the future (for the forward sum) and M years in the past (for the back sums), where I can toggle N and M by myself.

elbarto · ‎06-09-2022

I am using psmatch to do a simple propensity score matching as follows: A snapshot of my data is the following. DATA have; input year firm_id place_id treatment var1 var2; DATALINES; 2000 47500 33019 0 0.026167367 0.118107848 2000 81175 31079 0 0.021634581 0.11606627 2000 174572 34003 0 0.008156653 0.108517274 2000 174572 34005 0 0.007877977 0.091809787 2000 174572 34007 0 0.010265773 0.085315727 2000 174572 34015 0 0.010941376 0.086731717 2000 223005 9009 0 0.009077159 0.088563122 2000 268006 33013 0 0.01309349 0.101272136 2000 268006 50027 0 0.009887925 0.098948561 2000 174572 9009 1 0.008821509 0.104278341 2000 174572 33013 1 0.009961594 0.120187096 2000 174572 45234 1 0.004693746 0.119641706 2000 223005 31079 1 0.005699048 0.144180074 2000 223005 33019 1 0.005843101 0.112463228 2000 223005 42311 1 0.005326652 0.138652623 ; RUN; I want to do one-to-one nearest neighbor with replacement propensity score matching between treated observations (i.e., treatment=1) with control observations (i.e., treatment=0). Observations must be matched within the same year BUT each treated observation must be matched to a control observation with a DIFFERENT firm_id (that has the closest propensity score). I am using the following code: proc psmatch data=have region=allobs; class treatment year; psmodel treatment(Treated='1') = var1 var2; match method=replace(k=1) stat=ps exact=year caliper=.; output out(obs=match)=have_match matchid=_MatchID; run; data have_match(keep=year firm_id place_id treatment _PS_ _MATCHWGT_ _matchid);set have_match;run; However, in the resultant "have_match" dataset: we can clearly see that row 4 and row 5 are matched with row 1, which has the same firm_id of 174572. All other matches are fine because they are matched to a different firm_id. How can I fix the code using psmatch so that I can restrict the matches to a different firm_id? Thank you!

elbarto · ‎06-08-2022

I am using psmatch to do a simple propensity score matching as follows: A snapshot of my data is the following. DATA have; input year firm_id place_id treatment var1 var2; DATALINES; 2000 47500 33019 0 0.026167367 0.118107848 2000 81175 31079 0 0.021634581 0.11606627 2000 174572 34003 0 0.008156653 0.108517274 2000 174572 34005 0 0.007877977 0.091809787 2000 174572 34007 0 0.010265773 0.085315727 2000 174572 34015 0 0.010941376 0.086731717 2000 223005 9009 0 0.009077159 0.088563122 2000 268006 33013 0 0.01309349 0.101272136 2000 268006 50027 0 0.009887925 0.098948561 2000 174572 9009 1 0.008821509 0.104278341 2000 174572 33013 1 0.009961594 0.120187096 2000 174572 45234 1 0.004693746 0.119641706 2000 223005 31079 1 0.005699048 0.144180074 2000 223005 33019 1 0.005843101 0.112463228 2000 223005 42311 1 0.005326652 0.138652623 ; RUN; I want to do one-to-one nearest neighbor with replacement propensity score matching between treated observations (i.e., treatment=1) with control observations (i.e., treatment=0). Observations must be matched within the same year BUT each treated observation must be matched to a control observation with a DIFFERENT firm_id (that has the closest propensity score). I am using the following code: proc psmatch data=have region=allobs; class treatment year; psmodel treatment(Treated='1') = var1 var2; match method=replace(k=1) stat=ps exact=year caliper=.; output out(obs=match)=have_match matchid=_MatchID; run; data have_match(keep=year firm_id place_id treatment _PS_ _MATCHWGT_ _matchid); set have_match; run; However, in the resultant "have_match" dataset: we can clearly see that row 4 and row 5 are matched with row 1, which has the same firm_id of 174572. All other matches are fine because they are matched to a different firm_id. How can I fix the code using psmatch so that I can restrict the matches to a different firm_id? Thank you!

elbarto · ‎04-03-2022

Ah nevermind, I figured it out by writing a macro, thanks again!

elbarto · ‎04-03-2022

Thanks! That's really smart and efficient way of doing it. Do you have any ideas how to accomplish the last part? I would like to export each dataset to csv without the county_id variable. So for example, for Low_2002_tri_1, I would do: data Low_2002_tri_1(drop=county_id); set Low_2002_tri_1; run; proc export data=Low_2002_tri_1 outfile="E:\folder\Low_2002_tri_1.csv" dbms=csv replace; putnames=no; run; But, how can I loop this for all datasets?

Online Status	Offline
Date Last Visited	‎08-03-2024 05:02 PM

How to choose different random sample sizes using proc surveyselect?

How to transpose this table?

Re: Creating a timeseries based on information in a dataset

Creating a timeseries based on information in a dataset

Re: How to find a given word in a string

Re: How to find a given word in a string

Re: How to find a given word in a string

How to find a given word in a string

Creating a variable that counts consecutive years

Re: Forward and backward sums

Re: How to transpose and rename variables

How to find a given word in a string

Replacing values and outputting data as separate csv file

Assign group numbers to nearby ID's

How to choose different random sample sizes using proc surveyselect?

How to transpose this table?

Re: Creating a timeseries based on information in a dataset

Creating a timeseries based on information in a dataset

Re: How to find a given word in a string

Re: How to find a given word in a string

Re: How to find a given word in a string

How to find a given word in a string

Creating a variable that counts consecutive years

Re: Forward and backward sums

Forward and backward sums

Propensity score matching using psmatch - restricting matches

Propensity score matching using psmatch - restricting matches

Re: Replacing values and outputting data as separate csv file

Re: Replacing values and outputting data as separate csv file