Sorry for the crappy scan. Here’s the footer in case the header didn’t give it away.

What I want to do is delete everything and install only 2010. But sometimes I need 2003 for support reasons. So maybe I have 2010 and 2003.

I haven’t totally embraced the Virtual Machine paradigm for production work. I use it for beta testing or for specific OS/Excel needs (like when I’m editing a book), but I don’t use it for every day client work. I know some people do and I’m genuinely impressed by them. When Windows or Office has an update, do they go into all their VMs and run update? That sounds almost as fun as poking myself in the eye.

So what should I do? Your comments are always appreciated.

Update

This isn’t really an update, but is a question that is somewhat related. My laptop is nearing the end of its life. What if…stay with me here…I replaced it with a Macbook Air? I only use it about 20% of the time and mostly while traveling. I wouldn’t mind having a Mac for Rails stuff, although I have it working under Windows pretty well now. I’ve heard many times, and twice in the last week, someone say something like “I bought a Mac in 2006 and never looked back.” It makes me we want to see if I’m missing anything. However, my first love is Excel, so if it doesn’t do that well, I’m not interested. I expect an opinion or two on this.

Graph any Excel formula in a XY (Cartesian) chart or a Polar plot

In my case, for a long, long time I have wanted to draw any graph on a web page and do so easily. Some years back Google introduced an API that I experimented with but it never caught my fancy.

So, here’s my attempt with JavaScript and MS Excel. Of course, Microsoft has a less than stellar history on web-based Excel/Office products (Office Web Components comes to mind). But, maybe, this time around it will be different, given the push that the company as a whole is making for web-centric products.

Graph any Excel formula as a function of a single variable

I will post a “how I did it” article in a day or two together with links to whatever documentation / samples I could find on the ‘Net.

 
The format for the name is &fracnd; where n is the numerator and d is the denominator. Thus ½ is a half, and ⅞ is seven-eighths. The HTML code representations for these are:

 
In theory (more on “in practice” later) every representation in a row is equivalent. This is our test table to export to Wiki or HTML format:

 
Cell D1: =CHAR(32)&ROW()&”/”&COLUMN()-3, then fill down and right. The right hand column has a few test cases. I used that formulaic construction to keep Excel from doing the divisions. The Excel fractional format is “# ?/?” for single digit denominators. The intermediary space is important. It indicates that a fraction may follow, just as a forward slash indicates a fraction may be present. And a format of /?? is a fraction not translatable into HTML5. Turning these patterns into VBA, this is my MakeFracs() function. It checks that there is a slash, then that there is not a slash–digit–digit pattern, and finally that there is a “space–digits 1 through 7–slash–digits 2, 3, 4, 5, 6, 8″ pattern to screen out sevenths and ninths. If all of those pass, it substitutes in the &fracnd; formulation for the fraction.

 
At least that’s all I wanted it to do. In practice, Wikipedia and WordPress seem to be not fully onboard with HTML5 and do not handle all fifteen &fracnd; formats (I confirmed Firefox does). That added “j-loop” in the middle translates the &fracnd;’s into Dec code. This works fine, though it’s a step back from HTML5. Option is given to use Hex if desired. Your Excel table then looks like this:

 
Where N1: =MakeFracs(D1) filled down and right. Arranged that way you can see the HTML5 design thoughts. The very ugly website would look like this:

 
Frankly, I’m not sure that this is an improvement. You’ll come across Wikipedia editors, however, who are convinced of it. I thought about adding a trailing space, as in MakeFracs = VBA.Replace(Arg, sOUT, sIN & Chr(32)), but for every time I wanted to, I thought of an example where I didn’t, and the logic got very convoluted. Better I decided to put the space in the table where wanted and not in the function. My HTML tablemaker is here, but it’s being overcome by the hard steady march of technology. Wiki and CSS tablemakers are coming up. I used MakeFracs() in the above. No fractions were harmed in the making of this post.

…mrt
©¿©¬

In A2:A7 enter the values:
6494.55
1311.36
198.08
124.9
131.81
131.81

In A1 sum the values with =SUM(A2:A7)

In A9 compute the same result by double counting all the values and subtracting the total, i.e., =SUM(A1:A7)-A1

In A10 check if A1 and A9 are the same. =A9=A1. The result is FALSE, i.e., they are not the same.

8392.51
6494.55
1311.36
198.08
124.9
131.81
131.81

8392.51
FALSE

Turns out that A9 has a 1 in the 11th decimal place. 8392.510000000010

Fair enough. There’s the 15 digits of accuracy thing going on and we have a rounding error.

Now, build the same model except this time put the A1 formula in A8 (and adjust A9 and A10 accordingly). The result is A10 shows TRUE!

6494.55
1311.36
198.08
124.9
131.81
131.81
8392.51
8392.51
TRUE

As far as rounding errors go, there isn’t one. This time A9 contains 8392.510000000000

Like I wrote at the beginning. I don’t know why the two models yield different results.

So, earlier this week I analyzed the performance of my Adsense account and concluded that it no longer made sense to show these ads.  This post discusses my analysis and the role of Excel in it.

The data from Adsense system included, on a month-by-month basis, information on number of ads shown, number of clicks, and revenue (for me, not Google).  It also included derived information such as the revenue per click, called Cost-per-click, or CPC, and the conversion rate (clicks/ads shown).  Of course the last two are easily computed from the first three data items.

Before going further, it might help to understand the different drivers of Adsense revenue.  I thought of 3 key elements:

1) The kind of ad Google shows.  It has to be relevant to the website visitor.

2) The number of times a website visitor clicks on an ad, and

3) The amount that Google pays me for the click.

Interestingly enough, the website owner has no control over any of those elements!

OK, back to the Adsense data.  Google Adsense has an option to download the data in CSV format.  So, after downloading into Excel, I “normalized” the data so that all the numbers for January 2006 were 100.  This had two effects: one could look at large numbers and small numbers in the same visual space, thus making trend comparison easier, and it masked the actual data.

[A technical note.  The month column included the start and end date of each month, e.g., 2006-01-01 - 2006-01-31.  So, I added a new column to get a month as Excel understands it, using the formula =DATEVALUE(LEFT(B4,FIND(" ",B4)-1)), and formatted it to show only yyyy-mm.]

Next, I created a PivotTable and PivotChart showing on a month-by-month basis the number of views, clicks, and revenue over time.  Since Google provided data for two products (Adsense for Content and Adsense for Search), I filtered the PT to show only Adsense for Content.

As Figure 1 shows, the views (in blue) have gone up while the clicks (in red) and revenue (in green) have dropped.  So, this makes it evident that item 2 above (number of clicks) has not fared well over the years.

Figure 1

Out of curiosity, I decided to check if the revenues-per-click (my share of what the advertisers paid Google) had changed over the years.  As Figure 2 shows, item 3 in the list above (the amount Google paid me per click) remained steady until early 2009, dropped in 2009 and the first half of 2010, rose in the latter half of 2010 to 2.5 times the 2006 level and remained steady through 2011.

Of course, since the total revenue continued to drop, it must mean that the drop in the number of clicks was far greater than the increase in the CPC.

Figure 2

To test item 1 in the list above (relevance of ads), I checked 3 pages at random.  The results were, frankly, surprising.  Given Google’s reputation of delivering accurate search results, I would expect website ads to be related to the website content.  That, amazingly enough, was not the case.

On a page that explains supply and demand curves, there were ads for power tools and plastic enclosures.

Figure 3

On the TM AutoChart page there were ads for “Find Autos Near You” and project management software.

Figure 4

And, on the TM Randomize Slideshow page there were ads for Google Chrome, Microsoft Private Cloud, reverse look up of cell phone numbers, and a network monitor.

Figure 5

I am sure the relevance of the ads to the website content must make sense to Google’s Adsense algorithms but it surely escapes me.

In any case, I decided to pull the Adsense ads, at least temporarily.

Excel turns them into this.

January 2005 becomes January 5th of the current year. February 2005 becomes February 5th. That’s no good. In a spare row, I used this formula

to create the date I wanted. Then I copied and pasted special – values over the top of the original dates. It adds 2000 the day to get the correct year. Adding 1 to the month and taking the zeroth day will get you the last day of the month.

January 5, 2012 becomes January 31, 2005.