Calling all statistical / scientific experts...

[thefreeprisoner]

I have been keeping a daily record of my symptoms and the treatments I have been trying out.
I have tabulated everything in a spreadsheet which is available here in MS Excel 2003/2007 format:

Excel download(s)

Notes:

• I have recorded most of my symptoms using a scale of 0-5 where 0=Severe and 5=No symptoms.
• I have also tracked hours of sleep per night, and energy on a scale of 0-10 where 0=immobile and 10=back to normal.
• I have been trying Zantac (ranitidine) and Paracetamol(acetaminophen). 1 = 1 pill (for Paracetamol, that's 500mg).
• During this time I have gone on an elimination diet (I have recorded LL in the notes for the days where I started this and RTM1-8 for the weeks where I gradually reintroduced foods known to be troublesome in stages -- let me know if you want any more details about which foods, if you think there might be a correlation there.)

I tried to make a chart, but it's difficult to find correlations.
Can anybody have a look at this and see if you can spot any correlations between my treatments and symptoms?

Thanks SO much for your help.

Rachel xx

 

[Jenny]

I have been keeping a daily record of my symptoms and the treatments I have been trying out.
I have tabulated everything in a spreadsheet which is available here in MS Excel 2003/2007 format:

Excel download(s)

Notes:

• I have recorded most of my symptoms using a scale of 0-5 where 0=Severe and 5=No symptoms.
• I have also tracked hours of sleep per night, and energy on a scale of 0-10 where 0=immobile and 10=back to normal.
• I have been trying Zantac (ranitidine) and Paracetamol(acetaminophen). 1 = 1 pill (for Paracetamol, that's 500mg).
• During this time I have gone on an elimination diet (I have recorded LL in the notes for the days where I started this and RTM1-8 for the weeks where I gradually reintroduced foods known to be troublesome in stages -- let me know if you want any more details about which foods, if you think there might be a correlation there.)

I tried to make a chart, but it's difficult to find correlations.
Can anybody have a look at this and see if you can spot any correlations between my treatments and symptoms?

Thanks SO much for your help.

Rachel xx

Hi Rachel

Can't read this I'm afraid - even in full screen it's too small for my eyesight

But it would be possible to correlate say hours of sleep with symptoms next day for example, and do a few other things. I could do this fairly easily but don't have the stats package loaded on my home computer I'm afraid.

Jenny

 

[thefreeprisoner]

Wow thanks for responding so quickly, Jenny; you're clearly the best person of all to help!
What you're probably looking at is a preview of the file. You'll need to click on the "more" button which appears when you move your mouse over the file. Then click on "download" and you can open it in Excel.
Alternatively, if you're using a PC, you can right-click on the file and choose download which will work just as well.

There are 2 versions; one for Excel 1997-2003 users and one for Excel 2007 users.

Rachel xx

 

[Jenny]

Thanks Rachel - I haven't used Excel for years!

I'll have to leave others to do the stats for you.

J

 

[Frank]

Rachel i might dissapoint you here. But i think there are a few factors that make finding correlations nearly impossible: subjective data, ME is often coming and going, multiple treatments, stress and other non included data. I know Cort has had the idea of creating a program to link treatments to scores. But this needs lots and lots of analysis and protocols..

 

[thefreeprisoner]

Yeah, I suspected as much.
I did try to keep myself to one treatment at a time, but eagerness to recover prompted me to try most of them all at once... and I think getting a cold and allergy problems has probably monkeyed with the data too, as well as over-exertion, which needs more of a physical test.
I did think of adding the Romberg Test to my records, because it's a pretty objective test you can do on yourself, but my results of that vary depending on what time of day it is.
...sigh...

Rachel xx

 

[PoetInSF]

I've done this in 2008/9 and also found that it was not easy. Some of the problems were:

- it was difficult to numerically price disparate activities (how much is doing dishes worth vs. walking 3 blocks?)
- variable delay of onset of symptoms anywhere from a few hours to 2 days
- too many independent variables

What I found more useful was a "controlled" experiment rather than passive recording of all things that I do. For example, if you are experimenting with drugs, keep all other variables (activities, other drugs, sleep) constant as much as you can and take/not take the drug for a few weeks. Likewise, try an activity while keeping others constant. (eg, rest 2 days, walk a mile at 45 min-mile pace and observe your symptoms for the next 48 hours).

In the end, all I found was:

- there is a strong correlation between rest/activities and symptoms
- exercise within my limit is a medicine while outside of my limit is a poison.
- it's not easy to predict which activities are safe/dangerous. They need to be experimented individually for type/intensity/duration/frequency.

You might find http://www.recoveryfromcfs.org/chapter10.htm useful. (I came across Bruce Campbell recently when someone posted the link here, and I found that his overall experience almost completely identical with mine.)

 

[Gerwyn]

there is something called ANOVA which can find correlations between multiple coexisting variables I would have to look up the methodology

 

[Jenny]

ANOVA (Analysis of Variance) is a test that compares means and variances of several treatment groups, not correlations. The aim is usally to see if different conditions or treatments produce different effects.

I agree that there are lots of confounding variables here - you could theoretically look at partial correlations or some kind of multiple regression analysis, but the data is all self report and subject to all sorts of biases. Using a sledgehammer to crack a nut.

But if you think you've got a fair amount of reliable data on hours of sleep and symptoms next day you could look at that using a simple correlation coefficient. (It would probably be just as good to just 'eyeball' your ratings though.)

Jenny

Jenny

 

[CBS]

Data

Hi Rachel,

As someone who has conducted a few ANOVA's (as well as a few ANCOVA"S, MANOVA's, MANCOVA's and even a few Time Series Analyses), an ANOVA with an N of one is not the way to go. I haven't looked at your spreadsheet but what your doing can be useful as long as it is kept in perspective (I'm doing something similar for myself right now).

You're going to have more meaningful data if you group activities by roughly similar amounts of energy expenditure (perhaps group all chores, errands, driving, napping, sleep, etc. and then keep track of the time in each category each day). As for subjective complaints (aches and pains of various sorts), I would suggest that there are a number of ways that these can be reported; % of the day with a Sx (symptom), Ave. intensity of Sx, and peak severity of a Sx (use a t least a 5 point scale, I recommend something with which you're familiar/comfortable; I use 1-10 as it is so commonly used in medical settings as a scale for pain). Continue to keep track of med changes (date and dose, etc.) and then graph the trends for each of the symptoms (which you may also want to group by bodily system - GI, neurological, etc.). I find that it's easiest to simply use a pen to mark on a line graph the date of a med change.

Again, with an N of one a lot fo the interpretation is going to be subjective but you could do a simple T-test for difference (say minutes in an activity or % of day with Sx, eg. headache) prior to and after a med change.

Hope this helps.

Shane

 

[Jenny]

If you're really interested in recording treatments and tracking symptoms regularly you might want to have a look at the website http://www.patientslikeme.com. There are hundreds of FM/CFS sufferers on there.

Most seem to have FM rather than ME, but it's a useful tool for monitoring 'progress' anyway. You get graphs where you can identify symptoms in relation to changes in treatments over time.

Someone very helpfully suggested this on another thread. I've joined and already got several helpful emails from the administrators.

Jenny

 

[Gerwyn]

Hi Rachel,

As someone who has conducted a few ANOVA's (as well as a few ANCOVA"S, MANOVA's, MANCOVA's and even a few Time Series Analyses), an ANOVA with an N of one is not the way to go. I haven't looked at your spreadsheet but what your doing can be useful as long as it is kept in perspective (I'm doing something similar for myself right now).

You're going to have more meaningful data if you group activities by roughly similar amounts of energy expenditure (perhaps group all chores, errands, driving, napping, sleep, etc. and then keep track of the time in each category each day). As for subjective complaints (aches and pains of various sorts), I would suggest that there are a number of ways that these can be reported; % of the day with a Sx (symptom), Ave. intensity of Sx, and peak severity of a Sx (use a t least a 5 point scale, I recommend something with which you're familiar/comfortable; I use 1-10 as it is so commonly used in medical settings as a scale for pain). Continue to keep track of med changes (date and dose, etc.) and then graph the trends for each of the symptoms (which you may also want to group by bodily system - GI, neurological, etc.). I find that it's easiest to simply use a pen to mark on a line graph the date of a med change.

Again, with an N of one a lot fo the interpretation is going to be subjective but you could do a simple T-test for difference (say minutes in an activity or % of day with Sx, eg. headache) prior to and after a med change.

Hope this helps.

Shane

sure i was thinking about the symptom variability

 

[Alesh]

Correlation between energy and brain fog

Hi Rachel,

I think it is a wonderful job what you have done.

I looked at your data from January. I tried to find the correlation between "Energy level" and "Brain fog". I suppose the bigger the number in "Energy level" means actually the higher energy and the bigger the number in "Brain fog" actually means worse feeling of brain fog.

I supposed that the two vectors, each having dimension of 35, i.e., 35 days would be negatively correlated but when I computed their correlation coefficient I obtained +0.385359. So I computed the Spearman rank correlation coefficient, which is specifically designed to detect correlation, even if it is not linear, and I obtained +0.472435.

It is a bit surprising to me since my brain fog tends to anticorrelate with my energy. But I understand it is too small sample to draw any conclusion and there is some degree of bias since things like "Brain fog" is difficult to measure objectively.

In the past I kept similar records: I used to walk through a fixed path with fixed length and I recorded the times needed to finish this task. And it always helped me when I was in a state of desperation to look back at my few years old records and see some albeit small improvement. I also wanted to measure the degree of my tinnitus objectively by measuring the intensity of a precisely defined sound that would just be able to mask the tinnitus but I didn't have the necessary equipment.

Generally I think some kind of quantitative diary is an excellent antidote against desperation caused by the apparent persistent chronicity of ME/CFS.

Alesh

P.S.: I love the British pronunciation of English even if I've never been to any English speaking country.