We are trying to keep accurate records and track progress. Just wondering what if any advice anyone has on trendlines. Logarithmic? other?

Exponential, with y-intercept set to pre-TSM level seems to be most used.

I think there is some evidence that the sort of bio-physical process at work may be exponential in nature.

Thanks PV.

Looks like I might need some help with this. Exponential trend-line is grayed out for this chart and I can't set the intercept. any thoughts?

Thanks
L

Do you have any negative numbers or zero's in your dataset? That will cause Excel to disable selecting an exponential trendline. I enter AF days as .001 units to get around this.

My workplace blocks most image hosts so I can't tell if there's a screenshot in your post.

Oh, ok - I tried changing the y axis to show a higher starting point than zero, but I didn't think about the data set. Duh. I'll try that now.

Here is what it looks like with exponential trendlines with the intercept set to pre-TSM levels

Hi RGs,

Are your X-axis days? That's not gonna work. Either sum things up by week and you get 26.5, 26.5, 22, 22.5, more or less flat with slight decline, or do something like 7 days trailing average. For fitting curves it makes sense to include R-squared and see if the curve makes any sense at all.

Cheers,
D

Looks more like it now. this is trailing 7 day trendline. Thanks drunkie.

Hi Roxy

I have had lengthy discussions with Bob3D about this (we're both nerdy engineers ). The exponential is the appropriate curve to use.

What I do is record my consumption each day in a spreadsheet in one particular column. Then in the adjacent column I add a small increment (0.01) to each value - so that the values in this new column are never zero. I plot this column.
It is very important that you set the intercept of the exponential to equal your consumption prior to TSM - otherwise you may get funny results.

Some things to look out for (based on my experiences):
a) You need about 20 or so points to get a decent fit (depending on how much your consumption varies)
b) very small (i.e. zero in reality, but we've added a tiny amount to the 'raw' values) values will tend to distort the fit you get - if you have a lot of them, in particular, it will tend to suggest that the end (i.e. cure) is nearer (or even past!) than it probably is. If you have relatively few (zero days) then the opposite will be the case.
c) in the early stages (first 4 or so months for me) when I had maybe 2 AF days per month, the daily graph gave me the most realistic looking fit (the weekly graph suggested it would take me several years to reach 'safe' levels). Now I'm at the point where the daily graph suggests I'm cured a few months back (because I'm having typically 3 AF days per week and reduced consumption on the others), but the weekly (and I suspect soon the monthly) graphs are looking more realistic (I'm now about 7 months into TSM and estimate something in the 1-3 month region for hitting my target levels (an average of 14 UK units or less per week - although I'll want a couple of months of this at least before I'm calling myself cured).