Since my Pennsylvania model worked so well (it gave Obama just under 46% of the vote) I ran the same kind of calculations for Indiana. What I did was estimate the vote percentage in each Indiana congressional district by comparing it to demographically similar counties immediately across the state border. In the case of Indiana I used Illinois and Ohio. For interior Indiana CDs I used districts in ohio with similar racial demography and percentage of urban population. I then estimated what the proportion of total primary turnout would be from each CD by comparing the vote totals for Democratic House candidates between districts in 2006. Results below the fold.
First, a disclaimer. I'm a strong Obama supporter and I've done volunteer work for him in Pennsylvania and New York. So I'm not particularly happy with what I've found. To wit, Indiana is going to be hard for Obama to win. My model comes up with a popular vote total of Clinton: 52.6% Obama 47.4%. Obama will need a truly massive turnout in CD1 (Chicagoland) and CD-7 (Indianapolis), as well as in the college towns (Bloomington, in CD-9, and South Bend in CD-2) to win this.
Here's more data.
District
CD-1 (Chicagoland) ...percentage of voters for Dem House candidates in 2006 (state total); 12.8%. Estimated vote: O 68, C 32. Comparison: Obama vote in Chicagoland/IL.
CD-2 (North-central IN, includes South Bend)...% Dem electorate in '06: 12.8%. Estimated vote: O 45, C 55. Comparison: NW Ohio counties + 5 points for Obama for South Bend and Chicago media markets.
CD-3 (northwest corner of IN) ...% Dem electorate: 9.8%. Estimated vote: O 40, C 60. Comparison: NW Ohio counties.
CD-4 (central Indiana, no state border, just W of Indianapolis) ... % Dem electorate: 8.3%. Estimated vote: O 40, C 60. Comparison: NW Ohio - Ohio CD-5.
CD-5 (central Indiana, no state border, just E of Indianapolis) ... 7.9 % Dem electorate. Estimated vote: O 40, C 60. Comparison: NW Ohio - Ohio CD-5.
CD-6 (central and southern portion of border with Ohio) ...9.5% Democratic electorate. Estimated vote O 40 C 60. Comparison: central and southern Ohio counties on Indiana border. A question. The SE corner of CD-6 impinges on Cinncinnati suburbs. Could this increase Obama's percentage a bit in this district?
CD-7 (Indianapolis): 9.2% Dem electorate. Estimated vote O 67 C 33. Calculated this by assuming Dem electorate mirrors population racial demographics, viz. 60% African-American, 30% white. Assigned Obama 90% of AA vote, 34% of white vote.
CD-8 (SW corner of state): 16.1% of Dem electorate. Estimated vote: O 40, C 60. comparison: extreme SE Illinois.
CD-9 (SE corner of state/ Kentucky border): 13.5 % of Dem electorate. Estimated vote: O 45, C 55 comparison: Ohio CD-8 - western Ohio + 5 points for Obama due to Bloomington/U. Indiana.
I'd appreciate all feedback from Kossacks who know Indiana well. For example, hot congressional races in 2006 might have raised turnout disproportionately in particular CDs.
At any rate, my guesstimate is that Hillary will win Indiana narrowly, by about 5 points. Enough to keep her in the race, but not of course to stop Obama's ultimate nomination.