Case-Control Matching | SPSS

[@GerardfromMS]

Literally the best tutorial of this on the internet! You are fantastic!

[@souffremercuriel]

Thank you so much you saved my life ! A lot of kiss from a french surgery resident

[@fayewachs]

My research assistants and I watch this at least once a week…. Thanks so much!

[@lizzylartey]

Glad it was helpful!

[@sheilab4157]

Excellent!!! I need to know this for my analysis. Much appreciated :)

[@lizzylartey]

That's awesome

[@wenzhang8710]

thank you very much for your video. well explained and organized.

[@lizzylartey]

Thank you.Glad it was helpful!

[@petermuller5949]

Great video-Thanks alot! Still had to watch it 5 times to get it right 😂. In my case the Group Indicator had to have the values 0 and 1. Otherwise 1 and 2 strangely did not work ..

[@maisondemanon_]

Thank you for your reaction! First I got an error, but when I used the values 0 and 1, it worked!

[@MichelleBorckardt]

Excellent! Thank you so much. This is amazing!

[@abdifatahfarahhamud6426]

I am totally suprised the way you have suffered to help us; infact god will help you Insha allah.

[@flowerpixel]

saved me at work ty

[@ryanphillips8671]

Thank you so much for making this video, you have helped me so much! Upvoted and subscribed.

[@lizzylartey]

I'm glad it helped!

[@Ash-be1nv]

Thank you very much!

[@quincymoonen6181]

Thank you for the tutorial! I do wonder: how do I match e.g. 1 on 3? (Thus, not 1 on 1 as shown here)

[@arielhasidim]

Just add more "Names for Match ID Variables" names with space between in the dialog (fourth input field)

[@lukaku10832]

Thank you so much!!!!

[@digantgupta4074]

Excellent presentation. I was wondering if it is possible for you to please share a copy of the corresponding SPSS dataset so that we can follow along? Thank you.

[@begumbilgin2255]

I have 38 cases in total. It always matches only the 30 of them. Might that be about my tolerance value? What should I do? I was wondering if you know how to determine the best tolerance value for the continuous variables? Thanks a lot in advance! :)

[@PreetaS15]

Thank you, can you take this further to show how we can do a match "1 to many"?

[@irisbrinkman2234]

Maybe an very off question but i am wondering why they sometimes use pseudo event dates for the controls?

[@drsaluto]

Thanks for the video! Is it acceptable to have tolerances of 7 for age? I have 13 cases but tolerances smaller than that can only match at most 10 cases. Thanks again

[@lizzylartey]

Technically you can make the tolerance for age any number. But then you have to consider that the standard deviation for age will be +/- seven years. This might be an [huge] issue for instance if you will be conducting any age specific analysis.

[@jaemoolee4665]

Thank you for the video! I am trying to do case-control matching on SPSS but for some reason, I am not getting any output on the two columns: match id and matchgroup variable... only dots on the rows. Just from reviewing my database, I should be getting enough matches. I get "Warnings 5.3" or "Warnings 5.2" for my output. Would you happen to know why this might be happening?

[@lizzylartey]

Thanks for watching and thanks for your question. Typically when there are dots in the two columns ( i.e. match group variable and match ID variables) it may mean that there are no matches based on the variables and the match tolerances selected. Additionally, if you have written the syntax for this you may want to double-check the code. Warning messages may show up as the example below. If it does review the line below the warning. It may provide you with some ideas about where the error is coming from. E.g. >Warning # >Invalid Number

[@jaemoolee4665]

@@lizzylartey Hello, thanks for the response. Just from visualization, I see there are enough cases to be matched on. Also, when I get Warning 5.3, there's no other explanation of why the error is happening...

[@domiknows9564]

@@jaemoolee4665 Hi, did you manage to solve this issue? I have the same problem and cannot find a solution.

[@lizzylartey]

It may depend on the version of SPSS. Try some of the solutions in the following IBM support thread: community.ibm.com/community/user/datascience/communities/community-home/digestviewer/viewthread?MessageKey=fb29e2f2-fafc-4f5a-8223-acb37c1d1b54&CommunityKey=886b6874-0fb1-402c-8243-c70ef8179a99&tab=digestviewer#bmfb29e2f2-fafc-4f5a-8223-acb37c1d1b54

[@maisondemanon_]

Dear Jae, I am just reading your message now and probably you already fixed it as it is a year ago, but maybe this helps: Did you already try to label the case control 0 and 1? I first labeled it with 1 and 2, but then I had the same problem as you. But when I used 0 and 1 for some magical reason it worked! Hope it will work out. Best wishes, Manon

[@cocnursingexam3762]

What about unmatched case control study mis?

[@codescooter9961]

can you please guide me about the tolerance rate I don't know how to adjust for binary data

[@PreetaS15]

What process does this function use to match cases? How is it different from the Propensity Score matching?

[@lizzylartey]

Sorry for the late reply. To the best of my knowledge, propensity score matching ( PSM) is a way in which a case and control can be matched. Both ways ( the way I showed in the video and PSM) can be used to minimize the selection bias in a case-control study. However, propensity score matching matches based on the propensity score ("conditional probability of receiving the treatment rather than the control given the observed covariates" (Rosenbaum and Rubin 1983)). The method calculates the score for each subject and then matches subjects with similar or the same scores.The way I showed in the video doesn't match based on any score, I would liken it to a filtering technique almost. It basically says find me everybody with these covariates but belongs to treatment or no treatment group. Below are some open source links to propensity score matching that you may find helpful. www.ncbi.nlm.nih.gov/pmc/articles/PMC3144483/ www.publichealth.columbia.edu/research/population-health-methods/propensity-score-analysis www.frontiersin.org/articles/10.3389/fphar.2019.00973/full

[@HuascarD]

@@lizzylartey i would love a tutorial on how to use the PSM tool here in SPSS. Do you have a video about it? Or do you know of someone with a video about it? Thanks

[@areggrigorian155]

are you able to do another video showing how to do 1:2 matching? the suggestions in the comments are not working. thanks!!

[@lizzylartey]

Hi Areg, sure! I'll be filming the video in the next week or so.

[@areggrigorian155]

@@lizzylartey Thank you! If you can please explain how to do the 1:2 matching and make it be exactly a 1:2 ratio, that would be awesome. Right now, I get 80 matches to 192 matches for explains so its not 1:2. Thanks!

[@ahlamalzharani8724]

Excuse me, I have one case match with three controls, how can I do that in sass

[@DoctorsGk]

can you please help me how to interpret the spss binary logistic result for case control? i think it is different from the way we interpret in cross sectional study?

[@yeissonrivero4533]

I love you, thanks..¡

[@wjliu6597]

how you can do a case control matching in 1case : 2 control ratio? thanks

[@lizzylartey]

A suggestion was made by @Ariel Hasidim in the comments. In the dialogue box "Names for Match ID variables" type in whatever you want the variable to be called, followed by a space and then another name for the match ID variables. This works as long as there are other available matches that match your criteria, so consider the match tolerance levels.

[@melissa-re6oi]

@@lizzylartey thank you for the video I'm wondering if I do 1:2 ratio and use the technique that you mentioned, I would have to increase the tolerance on the matching variables right? I didn't quite understand how to do that? So if I want year of birth to have a tolerance of 5 years then I would right the tolerance as "5" or "0.5"?

[@lizzylartey]

@@melissa-re6oi Hi Melissa, It would be 5

[@jenniferpineda]

what does it mean if your chi-square is significant? I used matching with replacement. I found that my dichotomous variables are nonsignificant but my continuous ones are. Please help!

[@lizzylartey]

Hi Jennifer, it likely means the differences in sample a and sample b are not by chance (i.e. the groups are difference based on those variables). Also chi-sqaured test are used for categorical variables (e.g. yes, no), if you are trying to find the significance of the contunuous variable try using a t-test. If you separate out the groups and run descriptives on it, do you notice any differences ( i.e. lets say you have a variable age, in group a the mean age is 30, vs group b the mean age is 25)?

[@anjumshaikh3270]

Hallo Madam, I would love to work under you and learn from you. I have recently completed my masters in public health from Germany. Looking forward to your response.

[@sharonerdrich]

the music is really annoying'.. just saying. i can't concentrate with that ..

[@Ulriiich23]

love your video but I wish you could speak out more. your words don't come out clearly

[@lizzylartey]

Thanks for the feedback