“For all of the improved practices and new ideas of the past decade, philanthropy remains an underperformer in achieving social outcomes. One cause of this may be that philanthropists are essentially unaccountable. Businesses have shareholders, politicians have constituents, and nonprofits have funders. In contrast, philanthropists are spending their own money subject to only minimal constraints on their judgment. Of course, foundations have boards, but the boards are often themselves the primary decision makers and, in any event, are not externally accountable. In theory, the media can play a watchdog role. But even if they had the data and capacity necessary to monitor outcomes, their audiences tend to be more interested in personal anecdotes, especially tales of malfeasance.”

Agree or disagree, it’s an important read for the research & evaluation aspect of working in the nonprofit world.

Sample Size Calculators

Estimating a sample size for your project can be confusing, complicated and frustrating.  When it come down to creating the final plan, it’s a good idea to get an expert’s opinion.  But while you are thinking things over there are some good online resources to get you started.

Sample Size Calculator offers both sample size and confidence interval calculation. While this tool does not offer many parameters to control the level of precision of the calculation, it includes easy-to-understand definitions of common statistical terminology.soft

Raosoft presents a sample size calculator that takes into account the margin of error, the confidence level, and the response distribution. It also offers to show side by side what the margin of error would be like with various sample sizes.

SampleCalc has the capacity to calculate both sample size and the margin of error. Among the sample size calculation tools listed here, this is the tool with the most control over the statistical parameters (therefore the most precise). It is also the only one available both in English and in French. SampleCalc takes into account the confidence level, the proportion for the calculation of the margin of error, the design effect value, the margin of error and the expected response rate.

DeffCalc estimates the design effect value due to sample stratification.

Know Your Lines

The Center for Urban Pedagogy is a working example of innovative use of design and data for good.  They’ve got kids and teens and teachers and artists and politicians working together on public policy and civic engagement project.  They will train you to do it too.

One of their coolest projects is “Know your Lines” about “Hot on the heels of the Census data dump, CUP’s new issue of Making Policy Public takes you behind the scenes of the largely invisible redistricting process in which politicians often get to choose their voters, instead of the other way around.”  An important example of the fact that the way you collect data essentially influences the results you get from the data.

The concept of collective impact was introduced in 2011 in the Stanford Social Innovation Review as “a highly structured collaborative effort that can achieve substantial impact on a large scale social problem…”

The basic idea involves forming a common agenda, developing a collective measurement system, implementing mutually reinforcing activities and staying in connection.  A fifth element important to the success is sometimes the presence of a backbone organization whose sole objective is the collective impact rather than any particular mandate.

I have worked with collective impact in several situations with growing success, such as with the Peel Children & Youth Initiative and the Bill & Melinda Gates Foundation.

I am enthusiastic about the idea.  Lots more resources and details are available through the FSG organization.

Weave (Web-based Analysis and Visualization Environment) is out now in its Beta 1.0 and it seems to have a lot of potential.  It’s free.  It’s available to anyone.  And it has a lot of levels – novice to expert.  One of my favourite aspects, since I do a lot of multilevel modeling is that is can work with and visualize nested data.

The site offers lots of demos and examples.  Even for the beginner level, there is a bit of set-up but its capacity seems worth the investment.

But it got me thinking about family meals and socioeconomic status in children.  There is a lot of research out there in support of family meals.  But it’s an interesting question of causality.

There’s research that confirms that in the US, family meals are an important opportunity to develop strong parent-child relationships and family connectedness.  And that participating in family meals can encourage positive behavior and improved physical and mental health.

Interestingly, the ChildTrends  databank includes tables that shows children in poorer families, and those whose parents had less education, are more likely to have meals together with their families than are children with wealthier or more educated parents. Table 1 here and Table 2 here.

Which direction is the causality?  What measure is causing what?  Do healthier kids eat more family meals or do more family meals cause healthier kids?  And are the effects different in different socioeconomic families?  In my interest I searched down the Raw data from National Survey of children’s health.  This led to the amazing finding that one can order datasets, cleaned an free and ready in 5 business days from the NSCH.  I’ve ordered the data and will play with it for a while to see.

For non-profits, two of the most important “10 things you will learn in this book” are: “looks can be everything” and “It’s (still) all about stories”.  As I work with non-profit organizations to understand their data and to disseminate their results, these two ideas continue to play centre stage.  Standing in the way of effectiveness and impact is the divide between the people who understand how to analyze complex data and those who want to understand what that data says. The more I distill from experience, the clearer I see the two things that can close that gap:

1. Good design.  Use types of plots and charts that are charismatic and intuitive, not necessarily the most complex or technically correct.  Pay more attention to colours.  The psychology of learning shows us that this really matters. And don’t make stuff up.

2. Carefully considered communication style.  If it’s a complex analysis, tell it as a story.  A multilevel regression output is mind-numbing to most people.  The same results illustrated through a scenario can draw people’s attention and understanding and actually change policy.  A new piece of preliminary data, however, might work best as a simple percentage or number in a sentence.  Form needs to follow function.  After all, a good piece of statistical communication is often an affectional ordering of results.

Robert F. Kennedy made the point during his 1968 campaign;” When one moves beyond short-term ups and downs, though, things get more complicated. “Our gross national product…counts air pollution and cigarette advertising and ambulances to clear our highways of carnage, It counts special locks for our doors and the jails for the people who break them. It counts the destruction of the redwood and the loss of our natural wonder in chaotic sprawl.…Yet the gross national product does not allow for the health of our children, the quality of their education, or the joy of their play.”