I have a subtype of IBS which gives me mostly abdominal pain (IBS-U).
I have been keeping track daily of a health quality index (HQI) (think of it as a likert scale), what meds i am taking and how much, bristol stool scores (i track with an app), weight, how many minutes of cardio or weight lifting.
I have all the data imported into pandas and was wondering if someone could double check my work. I am performing a several single linear regression with each input and the HQI, e.g. "HQI ~ nexium’, ‘HQI ~ cardio_minutes’ etc.
Was wondering if anyone has used pandas here and can double check my work to make sure they get the same p-values. I am also looking for feedback on the best way to deal with effect delay in the time series. Do i simply shift the HQI for lots of steps and do the regression, or do a moving average of the HQI and regress on that? I am happy to share data also.
Thanks.
1 Like
Hi Isaac,
(1) Great work. I believe that collecting your data is a great first step and think it’s great you are pursuing data science to find answers.
(2) I am about to hire a biostatistician for another project. As soon as they are on board, we can ask them to sort out your data questions.
(3) In the meantime, have you tried Cross Validated on StackOverflow? There are thousands of statisticians who answer questions like yours. Read the FAQ though - you need to ask questions in specific ways for the community to answer your question.
(4) Have you considered your microbiome as well? Microbiomes are being found to be the source of a variety of conditions (i.e. just last week, hypertension was caused and cured in rats by swapping microbiomes.) It is virtually certain that the microbiome is related in some significant way to IBS.
(4a) Evidence might be seen in your data if you have seen a change in your symptoms while on antibiotics. Any chance you have that data?
(4b) Additional evidence might be seen if you take probiotics after being on antibiotics (i.e. cleanse then colonize). Any chance that is part of your dataset?
(5) We have a system for instantiating N=1 studies in a way that makes the entire process an assembly line that can be repeated. It could be interesting for us to wire up your data to the system and see if we can help in your analysis. You could even integrate your Pandas work into the assembly line if it turns out to be valuable.
Feel free to email if you are interested in chatting more. (gillam at gmail)
Best,
-Mike
Michael Gillam, MD, FACEP
Hi Mike,
I am a researcher by day so I am familiar with cross validated. I am actually meeting with my universities statistical consulting center today to discuss.
wrt 4, yes I have and this is where my research started. I have 13 samples taken over 18 months from ubiome. They analysis is here: https://github.com/isaacgerg/ubiome_longitudinal_analysis.
My analysis of my microbiome shows that its largely independent of my IBS. I believe it is related to nerve problems resulting from intense trauma from gallbladder disease (acalculus, just like my father). You can read more about that here: https://blog.picnichealth.com/complete-medical-records-help-scientist-reach-a-diagnosis/
wrt 4a, I also have taken probiotics for about 24 months now and they also do not show any sort of statistical significance. The biggest help is nexium and my doctors are not sure why. The hypothesis is that my stomach makes too much acid which makes it to the bowel and causes pain.
wrt 4b. Before my gallbladder disease was known, SIBO was suspected so I was put on the cedar sinia protocol (xifaxian + neomycin). No help there either because that wasnt the problem. I have all the data you suggest.
wrt 5, I would happy to work with someone about this. I will send you an email. If you can send me back proof of MD, i can give you access to my picnic health account and you can literally see every test that has been done.
Thanks for the response. Will see you in email!