RYM - Ryman Healthcare

[Vaygor1]

…..

Oh and because they pay a dividend greater than the operating profit they have to borrow money for that!
…..

RYM would disagree with that.

Refer http://www.sharetrader.co.nz/showthr...l=1#post481023

[winner69]

Vaygor, you might be interested in this plot of RYM earnings (reported eps) and shareprice over the years (as it happens the dots are in year order from left to right so joining the dots gives a shareprice line - well almost except for the year earnings went backwards).

The line of best fit is also shown which is the long term correlation between the 2.

I made a note that 2008 and 2011 were the years to buy for excessive returns for a long term holder and the last 2 years indicate future below average returns are likely

Not noted but 2006 wasn't a good time for an investor to buy if you didn't want to endure the paint of seeing your wealth shrinking or not growing for many years - but OK if you did a couta and held in there for the good times

[Vaygor1]

Vaygor, you might be interested in this plot of RYM earnings (reported eps) and shareprice over the years (as it happens the dots are in year order from left to right so joining the dots gives a shareprice line - well almost except for the year earnings went backwards).

The line of best fit is also shown which is the long term correlation between the 2.

I made a note that 2008 and 2011 were the years to buy for excessive returns for a long term holder and the last 2 years indicate future below average returns are likely

Not noted but 2006 wasn't a good time for an investor to buy if you didn't want to endure the paint of seeing your wealth shrinking or not growing for many years - but OK if you did a couta and held in there for the good times

Well noted thanks Winner & very interesting.
I guess not too many of these graphs on the NZX have the dots in order going nicely from left to right over the passage of time.

One question I have is that the line of best fit seems too low, I would have thought the line of best fit (only possible for this graph as the dots are in time order) is where the area bounded by the curve above the best fit line equals the area bounded by the curve below the best fit line. ie The orange line below.

Attachment 6336

Attachment 6337

I do understand though that this could be a false measurement as time is not actually a factor in this type of chart, so perhaps a better best-fit line is such that the total of each of the vertical distances from the best-fit line to each dot below equals the total of the vertical distances from the best fit line to each dot above. Just eye-balling it, this would bring the slope of the orange line down a bit I think, but not by much. Not sure this is right either but makes more sense than looking at areas.

[Vaygor1]

I've run enough regressions to say with (almost absolute) confidence that this is NOT the line of best fit. It would definitely be steeper. Those dots around the $8 mark are what we describe as "influential" and have an inordinate impact on the shape and location of least squares trendlines. Then again, perhaps you are using a different objective function (rather than minimising the sum of squared residuals)

Not sure but I think the x-axis may need to be logarithmic. You can see the data points getting further apart (maybe due to market variance?) along the horizontal plane. The graph is more like a scatter-plot of two non-time dependant variables. Interesting for sure.

[winner69]

NG, as you say about yourself sometimes I too was rather loose with my words. I meant to say the line of best fit based on 2002 to 2012 data extended out to 2014 with the assumption that 202/12 was more normal behaviour and 2013/14 has seen a year or two of behaviour not in line with previous times. Very remiss of me and probably immoral and unethical behaviour to disclose.

Highlights how far away from 'normal behaviour' the last 2 years have been for the RYM shareprice.

Pace car out at Bathhurst again

[Goldstein]

Well noted thanks Winner & very interesting.
I guess not too many of these graphs on the NZX have the dots in order going nicely from left to right over the passage of time.

One question I have is that the line of best fit seems too low, I would have thought the line of best fit (only possible for this graph as the dots are in time order) is where the area bounded by the curve above the best fit line equals the area bounded by the curve below the best fit line. ie The orange line below.

Attachment 6336

Attachment 6337

I do understand though that this could be a false measurement as time is not actually a factor in this type of chart, so perhaps a better best-fit line is such that the total of each of the vertical distances from the best-fit line to each dot below equals the total of the vertical distances from the best fit line to each dot above. Just eye-balling it, this would bring the slope of the orange line down a bit I think, but not by much. Not sure this is right either but makes more sense than looking at areas.

As others have mentioned best to use the least squares fit.

Note that the slope of the resulting (red) line is the PE ratio. There is a very good fit up to and including 2012. There's quite some evidence though that the PE ratio changed during 2013. If you use only the 2013 and 2014 data points (and constrain it to go through the origin) you end up with something a lot steeper and more like the orange line.

Question: Which one is currently the one to use (i.e., are we going to revert to the historical PE ratio)?

[Vaygor1]

Yup. It's effectively a cross sectional correlation analysis of time series data. Totally legit if done properly, but also highly prone to spurious inferences due to "stationarity" problems.

You are welcome to correct me if I am wrong, but time forms no axis on this chart so using traditional methods is questionable. Yes, each dot represents a data point at a given instance in time at regular intervals once a year, but if RYM experience a particularly bad year (say this financial year) then the 2015 red dot would almost certainly appear in the bottom left quadrant of the chart somewhere, and Time (for this chart) is well and truly out the window.

Using a bounded area to look at an the average cannot normally be used in this type of chart at all, but as qualified when I posted it, it can (kind of… just) be used on this occasion because as it happens the years 2002-2014 line up left to right nicely.

As such, I'm not sure what the right approach is. I think a least squares fit is up for debate too.

I am currently thinking about it but like any statistical analysis, differing methods and views can lead to differing results; none of them necessarily right or wrong.

[Goldstein]

Yes, you have a scatter plot there with a line drawn say using least squares fit. This procedure assumes each point is completely independent, order doesn't matter etc. No matter how you order those points you'll still get the same line.

However, by ordering the points in time you start to see something interesting, which you do here.

If you start to worry about relationships between points from year to year (serial correlation) , then you lose the independence between points and you need to so something a lot more involved as NewGuy suggests.

[Vaygor1]

So I did a bit of work. Tossed various scenarios here and there and decided to throw it away as a Log/Log seemed worth a try.

It occurred to me that as both the EPS and Shareprice numbers (ie both x and y axis) increased from 2002 to 2014 that variance increased too. This was evident in the bunching up of the data points of both the EPS and the share price in the earlier years, towards the left and the bottom respectively on the linear chart.

I further realised that in terms of percentage difference between any data point and its corresponding value on a a variety of lines-of-best-fit on a linear plot that the 2013 data-point was not as far off the mark as either 2006 or 2009. So on this basis if one opted to treat 2013 as an out-lie which seems a reasonable thing to do then 2006 and 2009 should be treated as out-lies too.

Anyway, with no trendline, here is the result. I have labelled each data point with its corresponding year:

Attachment 6345

Now I have never taught stats, but I did get 97.5% in it at 2nd Pro Engineering school at Canturbury Uni (never was able to figure out at the time where I lost that elusive 2.5% - sorry, can't find a smily face for a nerd) and it was all too long ago with little application since to re-educate myself on it. But throwing some trendlines in there, using excel, bears some interesting results. Any advise before I post some trendlines? On excel, a linear trendline is not a straight line (but nearly) whereby a power trendline does in fact produce a straight line on a log/log scale and seems to be the best one to use as RYM's figures generally speaking have been increasing at a fairly constant rate to-date.

[Vaygor1]

Nice graph Vagygor. I am not surprised that log/log transformation reveals a linear relationship. This is tantamount to a constant elasticity between EPS and Price. In other words, a given % change in EPS is associated with a constant % change in Price. This is to be expected IMHO.

Sorry about the 2.5%. I was never happy to leave any marks on the table

Thanks NG.

While what you say is true and correct, as you no-doubt know the SP-EPS relationship will differ between different types of business. ie A software company will have a very different SP-EPS relationship than a power station or an architectural consultancy etc.

This exercise is proving to be very insightful, thanks for sparking the interest up Winner… and such a relationship between the EPS and SP could well be used to analyse other companies in the same sector (ie SUM and MET) to give investors an additional method of arriving at a future (rational market) share price against an estimated future, or recent announced, EPS.

Arrrgh…. now I know where my 2.5% went !!!

Will post some trendlines later.